Managerial Incentives, Financial Performance, and Governance Quality:

Evidence from Nonprofit Organizations

Ashley N. Newton

Division of Finance

Michael F. Price College of Business

University of Oklahoma

Norman, OK 73019-0450

anewton@ou.edu

September 2013

Abstract

Systematic differences in the organizational forms of nonprofit and for-profit entities have

elicited a variety of concerns that remain largely unaddressed. Specifically, the nonprofit sector

is characterized by weaker monitoring mechanisms and potentially more severe agency problems

relative to their for-profit counterparts, due in large part to an absence of shareholders in

nonprofit organizations. These unique aspects of the nonprofit organizational structure result in

a heightened sensitivity about their governance quality, particularly with respect to its ability to

mitigate the undesirable consequences of agency conflicts, such as excessive executive pay and

poor financial performance. To address these concerns, I study the relationship between chief

executive compensation, financial performance, and governance quality. I find a significant

negative relationship between the CEO-to-employee relative pay ratio and multiple measures of

nonprofit financial performance. I document a similar inverse relationship between executive

perquisites and nonprofit performance. The results also highlight the importance of corporate

governance; specifically, strong governance mechanisms (1) mitigate excessive executive pay in

nonprofits marred by potentially severe agency conflicts, and (2) significantly negate the

decreases in performance that tend to accompany increases in relative pay. Collectively, these

results demonstrate a positive association between disparities in nonprofit executive pay and

agency conflicts, and highlight the importance of implementing strong governance mechanisms

as a means to curb self-interested managers’ tendencies.

I thank Louis Ederington for valuable comments and suggestions.

2

I. INTRODUCTION

Stark differences in the organizational forms of nonprofit and for-profit entities have

elicited an assortment of concerns about the role and responsibilities of a nonprofit manager.

However, and despite the prominence of the nonprofit sector in the U.S. economy – specifically,

encompassing 5.5% of U.S. GDP and 9% of the nation’s labor force (in 2010)1 – many of these

concerns remain unaddressed. Of particular importance is the nonprofit organization’s absence

of shareholders and the implications such lack of oversight has on the motives and actions of

nonprofit managers and boards of directors. Given that the managers and directors of nonprofits

are not held accountable to (and are not themselves) owners of the organization, nonprofit

entities are exposed to agency conflicts that are arguably more prevalent and of greater severity

than those facing for-profit organizations. These concerns draw special attention to nonprofit

governance quality, since implementing strong governance mechanisms should help to ward off

undesirable consequences that are likely to accompany agency conflicts, such as excessive

executive pay and poor financial performance. This paper presents evidence targeted at

answering four main research questions: (1) What factors determine CEO pay and, separately,

executive perquisites at nonprofits?, (2) What factors explain organizational performance?, (3)

What is the relationship (if any) between CEO pay and financial performance?, and (4) What

role does governance quality play with respect to the relations described in (1), (2), and (3)?

Relatively high compensation to a CEO could arise in either an optimal or sub-optimal

setting. For example, excessive CEO pay could originate in the case of a talented and very hard-

working CEO who operates in the best interests of the nonprofit, and thus who deserves to be

rewarded for his efforts (Pay-for-Performance Hypothesis). In contrast, abnormally high CEO

compensation may arise if the CEO is able to use his power, influence, and freedom from

1 See Roeger, Blackwood, and Pettijohn (2013).

3

monitoring to extract compensation in excess of what he deserves (Agency Hypothesis). While

the former example highlights a scenario of an optimally set (though high) level of CEO

compensation, the latter profiles a case of sub-optimally high CEO pay. Moreover, a sub-

optimally high level of CEO compensation is consistent with agency conflicts, whereby the CEO

prioritizes his own well-being rather than (and often at the cost of) that of his organization. To

disentangle these hypotheses, it is critically important to account for corporate governance

quality since governance quality should be correlated with both CEO pay and organizational

performance. As such, this paper tests whether the relationship between CEO compensation and

nonprofit financial performance exhibits either a positive (Pay-for-Performance Hypothesis) or

negative (Agency Hypothesis) association and how this relation depends on nonprofit

governance quality.

I investigate the intersection between CEO pay, financial performance, and governance

quality in the context of a large and diverse sample of nonprofit organizations during 2008-2009,

where financial data are drawn from the significantly redesigned Internal Revenue Service (IRS)

Form 990, Return of Organization Exempt from Income Tax, which was released in tax year

2008. By most accounts, the most notable additions to the new form were questions related to

governance quality – organizational characteristics that certainly have the potential to

significantly affect the relationship between executive pay and firm performance. The form

featured many additions including a section entitled, “Governance, Management, and

Disclosure”, touted as the revamped form’s “crown jewel” by Steven T. Miller, former

Commissioner of the Tax Exempt and Government Entities Division (Miller 2008). These new

data allow for a comprehensive analysis of nonprofit governance quality that was previously

impossible to conduct. A focus on governance quality addresses the agency conflicts plaguing

4

the nonprofit sector and, by extension, more clearly distinguishes between the Pay-for-

Performance and Agency hypotheses.

To explore these contrasting views of CEO compensation, I focus on the difference

between CEO and average employee compensation. Pay of both executives and non-executives

likely differs considerably from one nonprofit to another. For instance, salaries at hospitals or

major universities tend to exceed those in preschools and local theatre groups. To partially

account for this, I utilize a “relative pay ratio,” defined as (the natural log of) the ratio of CEO

compensation to average non-CEO employee compensation, where compensation encompasses

“reportable compensation from the organization” (essentially, that which is reported on a W-2).

Linking relative measures of CEO pay has become an increasingly popular practice in the for-

profit literature (e.g., Bebchuk, Cremers, and Peyer 2011; Faleye, Reis, and Venkateswaran

2013), particularly given their relevance to recent regulatory guidance and reporting

requirements (e.g., the Dodd-Frank Act). Nonprofit financial performance is captured through

four (industry-adjusted) measures, including the ratios of program expenses to total expenses,

program expenses to total assets, net fundraising revenue to total fundraising revenue, and (the

log of) total revenue per employee (Desai and Yetman 2006; Faleye et al. 2013). Governance

quality is defined as an index of numerous mechanisms that illustrate the efficacy of the

nonprofit’s governing body, governing policies, compensation policies, and accountability and

transparency.

I begin the analysis by examining the determinants of the CEO-to-employee relative pay

ratio. I find that relative pay is negatively related to financial performance, governance quality,

government revenue, and median household income, and is positively related to organizational

size and CEO tenure. The statistically significant negative association between the relative pay

5

ratio and nonprofit performance is robust to including several factors that may impact both the

relative pay ratio and financial performance, including size, donations growth, liquidity, and

capital expenditures. This result supports the view that both nonprofit CEO pay and financial

performance are subject to agency conflicts, i.e., that at poorly governed nonprofits,

insufficiently monitored CEOs are apparently able to extract excessive compensation while

minimizing their efforts.

Next, I model the determinants of four measures of (industry-adjusted) nonprofit

financial performance. Controlling for traditional determinants of performance, as well as

factors that may impact both the relative pay ratio and nonprofit performance, I again document

a significant and economically meaningful relation between the relative pay ratio and nonprofit

performance. For example, a single standard deviation increase in the relative pay ratio implies

approximately 1%, 7%, 1%, and 17% reductions in the (industry-adjusted) ratios of program

expenses to total expenses, program expenses to total assets, net fundraising revenue to total

fundraising revenue, and (the log of) total revenue per employee, respectively, relative to their

mean values. These results provide support to the Agency Hypothesis.

I address the likely endogeneity between CEO compensation and financial performance

using two-stage least squares procedures, a changes specification, and by modeling financial

performance as a function of the lagged (not contemporaneous) relative pay ratio. The empirical

results stand firm throughout these analyses, lending further support to the Agency Hypothesis.

I also investigate how other forms of managerial incentives relate to financial

performance and governance quality. Inferior financial performance may indicate ineffective

monitoring by the organization. Moreover, ineffective monitoring mechanisms tend to create

new or aggravate existing agency concerns, a result of which can be the over-consumption of

6

executive perquisites. Consistent with this theory, I find that weak governance quality (and

ineffective monitoring mechanisms) is associated with a higher likelihood of executive

perquisites, where perquisites include incentives such as first-class travel, discretionary spending

accounts, and housing allowances. Moreover, inferior financial performance (also suggestive of

ineffective monitoring mechanisms) is associated with a greater prevalence of executive

perquisites. Taking the opposing view, what can be said for an organization that is characterized

by strong governance quality and superior financial performance? I find that, although well-

governed and high-performing nonprofits are individually less likely to award CEOs with

perquisites, good governance quality is associated with an incremental but significant increase in

the likelihood of perquisites among high-performing entities. Collectively, these results suggest

that while the Agency Hypothesis is applicable to the average nonprofit organization, the Pay-

for-Performance Hypothesis is supported for nonprofits that are both well-governed and high-

performing.

This paper also presents evidence on patterns in CEO pay and nonprofit performance

surrounding CEO turnovers. While the results indicate higher attrition rates for excessively

compensated CEOs, the probability of a highly-paid CEO leaving his position is reduced if he

achieves superior financial performance in the year prior to turnover. Thus, although nonprofit

organizations (in general) do not appear to pay for financial performance, the probability of

attrition among highly-paid CEOs is lower when the organization is performing well.

As a more in-depth analysis of the Agency Hypothesis and its implications, I investigate

the relation between governance quality and the CEO-to-employee relative pay ratio for varying

degrees of agency conflicts. Superior governance quality should prove most valuable in

mitigating excessive executive pay when agency problems are severe. Approximating nonprofit

7

agency problems as a function of excess liquid asset levels (as in Core, Guay, and Verdi 2006), I

find that several governance mechanisms serve to alleviate excessive executive pay in

organizations characterized by potentially severe agency conflicts. For example, for nonprofits

marred by agency problems, the decision to adopt strong governing policies is associated with a

large reduction in relative pay levels. Specifically, choosing to adopt a conflict of interest policy

implies a 15% reduction in the relative pay ratio, while adopting a whistleblower policy is

associated with a 14% reduction in relative pay. These results highlight the economic

significance of governance quality. To further validate the value of governance, I investigate

whether good governance quality ensures that changes in the relative pay ratio are accompanied

by improvements in financial performance. The empirical results support this prediction,

suggesting that strong governance mechanisms promote better financial performance

surrounding changes in relative pay arrangements, such as when changes are made to a CEO’s

compensation contract.2

This paper contributes to extant literature of executive pay. Relevant studies in the

nonprofit sector find a positive link between executive pay and nonprofit size (e.g., Aggarwal et

al. 2012; Hallock 2002) that varies across industries and ownership types (Ballou and Weisbrod

2003). Much of the emerging literature of for-profit CEO compensation investigates how

relative measures of CEO pay relate to a variety of corporate outcomes (e.g., Bebchuk et al.

2011; Faleye et al. 2013). I advance research of nonprofit executive pay by considering a similar

2 I conduct several robustness tests to validate the results. First, I split the sample into “commercial” and

“traditional” sub-samples, where commercial nonprofits encompass those that tend to rely on a single source of

revenue (program services revenue). Second, given the importance of organizational size as a determinant of both

CEO pay and financial performance in the nonprofit sector, I re-estimate the results using different measures of nonprofit size. Third, I provide additional results that utilize alternative measures of CEO-to-employee relative

pay. Finally, given that some U.S. states mandate certain governance standards, I revisit the primary pay and

performance regression specifications after incorporating state fixed effects. Tabulated results, provided in

Appendix B, lend further support to the Agency Hypothesis and its prediction of a negative link between the

relative pay ratio and financial performance.

8

and timely definition of CEO compensation that is measured relative to the average employee’s

salary.

This paper is also related to existing studies that link executive compensation to firm

performance, most of which are drawn from the for-profit literature. In the for-profit realm, a

positive association between executive pay and firm performance has been documented on

numerous occasions (e.g., Jensen and Murphy 1990; Conyon and He 2011). From the much

smaller body of evidence of the nonprofit sector, Baber, Daniel, and Roberts (2002) find a

positive association between nonprofit executive pay and charitable spending, while Sedatole,

Swaney, Yetman, and Yetman (2012) find that the two are not significantly related. I expand

upon these largely inconclusive findings by relating a more timely definition of CEO pay to

multiple measures of nonprofit financial performance, finding a strong and consistent negative

relation.

This paper also contributes to existing studies of governance quality and its impact on

both executive pay and firm performance. An examination of the importance of corporate

governance mechanisms in nonprofit organizations is particularly essential, given the prevalence

of weak governance quality in nonprofits relative to their for-profit counterparts (Glaeser 2003).

Extant literature demonstrates that poor governance quality alienates donors (Hansmann 1980),

while improvements in governance are associated with more efficient operations (Alexander and

Lee 2006). Related studies highlight the importance of a nonprofit’s governing board, with

larger boards being associated with decreased monitoring incentives (O’Regan and Oster 2005),

and the inclusion of donors on boards leading to potential efficiency improvements (Callen,

Klein, and Tinkelman 2003). While these findings are informative, notable deficiencies warrant

a more in-depth and comprehensive analysis of nonprofit governance quality. Specifically,

9

extant literature fails to offer a definitive prediction as to how executive pay should relate to

nonprofit performance, particularly after controlling for governance quality. Moreover, recent

regulatory reforms (e.g., Dodd-Frank) highlight the relevance of correlating relative pay

measures with performance. Although neglected in the nonprofit literature, this objective has

received significant attention in the for-profit realm (e.g., Bebchuk et al. 2011; Faleye et al.

2013). Using governance data newly made available on IRS Form 990, this paper provides an

all-inclusive and enlightening account of the impact of nonprofit governance quality on both

managerial incentives and financial performance.

This study also adds to extant studies of CEO turnover. In the nonprofit realm, Brickley

and Van Horn (2002) find that better financial performance lowers the probability of CEO

attrition. Eldenberg, Hermalin, Weisbach, and Wosinska (2004) find a similar result in their

study of nonprofit hospitals. Empirical evidence of for-profit entities also supports an inverse

relationship between firm performance and the likelihood of CEO turnover (e.g., Jenter and

Kanaan 2013), while Kaplan and Minton (2006) clarify that board-driven (but not takeover- or

bankruptcy-driven) turnover is significantly related to firm performance. I advance this stream

of literature by examining both financial performance and governance implications surrounding

CEO departures in the nonprofit sector.

The next section presents a discussion of relevant literature and develops my hypotheses

regarding executive compensation, financial performance, and corporate governance quality.

Section III describes the sampling procedure. Section IV reports the results of the empirical

analyses, and section V presents several robustness tests. Conclusions are offered in section VI.

10

II. NONPROFIT COMPENSATION, PERFORMANCE, & GOVERNANCE

The nonprofit sector is a substantial and increasingly important component of the U.S.

economy. In 2012, there were 1.6 million nonprofit entities registered with the IRS. As of 2010,

the nonprofit sector encompassed 5.5% of U.S. GDP and 9% of the nation's labor force. At this

time, public charities reported an aggregate $1.5 trillion in revenues, $1.45 trillion in expenses,

and $2.71 trillion in total assets. Gifts from individual foundations, corporations, and bequests

reached approximately $290.89 billion. Interestingly, while many other sectors of the economy

struggled during the 2007-2009 financial crisis, most nonprofit industries thrived, reporting an

increase in both employment and wage levels in 2010 relative to 2000 on the order of 17% and

29%, respectively (inflation-adjusted). This positive pattern runs counter to that of the for-profit

business sector, which endured a drop in both employment and wage levels over the same period

of –6% and –1%, respectively (Roeger, Blackwood, and Pettijohn 2013). As of year 2010, the

median CEO pay among 282 large charities was $475,192 – a 2.7% increase over the prior year

(Gose and López-Rivera 2012).

II.1. Nonprofit compensation.

As a matter of law, nonprofit organizations “must not be organized or operated for the

benefit of private interests…” (IRC § 501(c)(3)(d)(ii)). Even so, numerous cases of excessive

nonprofit executive pay have been documented. For example, Philip Levy and Joel Levy

(brothers), two top executives of a Medicaid-financed nonprofit organization serving the

developmentally disabled in New York, faced significant scrutiny for earning nearly $1 million a

year, driving state-funded luxury cars, and passing their children’s tuition bills onto the nonprofit

group (Buettner 2011). In response, Governor Andrew Cuomo of New York created a task force

to investigate allegations of excessive nonprofit pay that ultimately culminated in new

11

regulations, among them a provision that prohibits state-funded providers from devoting more

than $199,000 in state funds towards a single executive’s pay (Cuomo 2012). As another

example, in mid-2013, Senator Charles Grassley – a long-time critic of nonprofit hospitals –

voiced serious concerns about nonprofit boards rewarding CEOs with large monetary bonuses in

return for their increasing the quantity, not the quality, of services provided (Hancock 2013).

One of the many other heavily scrutinized cases of apparent abuse of nonprofits’ tax-exempt

status is that of American Bureau of Shipping, a company advertising “to promote the security of

life & property on the seas” (American Bureau of Shipping 2011) for the past 150 years of its

existence as a nonprofit organization, and one that “has earned hundreds of millions of dollars in

the past decade; lavished its executives with multimillion-dollar pay packages and perks; and

purchased an offshore hedge fund [in the Cayman Islands]” (Evans 2012).3

While these and other well-documented cases of excessive executive compensation in

nonprofit organizations raise obvious concerns regarding the quality of governance in the

nonprofit sector, questions on the origins of variation in nonprofit executive pay schemes – both

within firms (relative to non-executives) and across firms (relative to competing organizations) –

have largely been ignored. That is, while existing studies of executive compensation within

nonprofit organizations are informative, they leave much to be desired. Relevant literature

consistently supports a positive association between nonprofit executive pay and organizational

size (e.g., Aggarwal et al. 2012; Frumkin and Keating 2010; Hallock 2002), systematic

differences in executive compensation across different industries and ownership types (Ballou

and Weisbrod 2003), and a positive association between excess executive pay and excess liquid

asset levels (Core, Guay, and Verdi 2006). However, the literature fails to offer conclusive

3 For other examples involving allegations of excessive nonprofit executive pay, see Dennison (2011), Doyle (2011),

Fries (2013), Kruesi (2011), Pogrebin and Taylor (2010), and Sataline (2010).

12

evidence on many important issues involving nonprofit compensation, such as how executive

pay relates to organizational performance and corporate governance. A discussion of relevant

literature relating executive pay to financial performance is presented in the next section, and a

review of studies of executive pay as it relates to governance quality follows in section II.3.

II.2. Executive pay and nonprofit performance.

Conventional economic theory predicts that more capable managers will be rewarded

with greater compensation, implying a positive association between pay and performance.

Specifically, and in the context of for-profit companies, managerial incentive contracts are

designed to reward CEOs for taking actions and making decisions that increase shareholder

wealth. The implication of a positive association between executive pay and financial

performance has been repeatedly documented among for-profit entities (e.g., Jensen and Murphy

1990; Conyon and He 2011). The type and style of CEOs have also been demonstrated to

meaningfully impact corporate outcomes (e.g., Bertrand and Schoar 2003; Malmendier and Tate

2009). Moreover, Ryan and Wiggins (2004), among others, point to the type – and not just

amount – of managerial incentives as being incrementally important in terms of its impact on

firm performance.

Relative to the for-profit sector, much less focus has been devoted to studying how

executive pay relates to performance in nonprofit organizations. Baber, Daniel, and Roberts

(2002) find that changes in executive compensation are positively related to changes in program

spending, while Sedatole, Swaney, Yetman, and Yetman (2012) find no significant relation

between executive pay and program spending. As for other financial metrics, Brickley and Van

Horn (2002) find that executive pay is positively related to return on assets. The inconclusive

nature of this evidence is due at least in part to a lack of consensus about how to appropriately

13

measure the performance of a nonprofit organization. As explicitly described by Roomkin and

Weisbrod (1999), it is much more difficult to identify the performance-related incentives of

nonprofit managers relative to those of their for-profit counterparts.

II.3. Governance quality and agency concerns.

A particularly important difference between the organizational forms of for-profit and

nonprofit entities lies in their ownership structures. Specifically, as noted by Jensen and

Meckling (1976), one of the primary avenues through which agency conflicts arise within for-

profits is between their agents (managers) and primary principals (shareholders), particularly

when management fails to fulfill its obligation to maximize shareholder wealth. In stark

contrast, nonprofit organizations do not have residual claimants. Instead, nonprofit managers

and boards of directors face only informal obligations to a stakeholder base consisting of donors,

physicians, non-managing employees, and the community they serve. As such, the agency

conflicts plaguing nonprofit entities are certainly starkly different and likely more serious than

those associated with for-profits.

The governance structures of nonprofits have been labeled as being much weaker relative

to their for-profit counterparts (e.g., Glaeser 2003). This concern is likely aggravated by the

nonprofit sector's non-distribution constraint, which specifies that nonprofits are prohibited from

distributing net earnings to any individual who exercises control over the organization. This

provision can mistakenly lead both internal (non-managing) and external stakeholders to believe

that strong corporate governance mechanisms are simply not necessary. Consequently, some

U.S. states have instituted mandatory governance standards (Desai and Yetman 2006).

Empirical evidence further validates the importance of governance quality in nonprofit

organizations, with Hansmann (1980) finding that poor governance quality alienates donors, and

14

Alexander and Lee (2006) demonstrating that improvements in governance quality can lead to

more optimal operating efficiencies. Regarding nonprofit boards, O’Regan and Oster (2005)

find that the size of a nonprofit’s board is of great importance, with larger boards being

associated with decreased monitoring activities. Relatedly, Callen, Klein, and Tinkelman (2003)

find that including donors on the governing board can translate into efficiency improvements.

An important distinction between the governance practices of nonprofit boards relative to

their for-profit counterparts is the structure and operations of their boards of directors. For

example, while in the for-profit realm shareholders possess the power to replace directors,

nonprofit organizations lack shareholders altogether. Although some nonprofit board regulations

vary by state, nonprofit organizations are typically allocated significant (if not total) discretion

over board member replacement decisions. As such, policies as outlined in bylaws will tend to

vary across organizations, with no uniform or external mandates imposed on the nonprofit sector

as a whole (Pakroo 2011). Fama and Jensen (1983) provide further details about nonprofit

boards, including an account of what nonprofit boards should entail: board members that are

primarily if not entirely external to the organization, most if not all of whom meaningfully

contribute to the organization through their wealth or time, and all of which are prepared to serve

without pay. Moreover, future members should be chosen from these independent, existing

members (to create a “self-perpetuating” board). Subsequent empirical evidence has found

larger boards to be associated with decreased monitoring activities (O’Regan and Oster 2005),

and that the inclusion of donors on nonprofit boards can translate into efficiency improvements

(Callen, Klein, and Tinkman 2005).

Governance quality also offers implications for executive compensation, yet only in the

for-profit literature has this relation been studied in-depth (e.g., Core, Holthausen, and Larcker

15

1999; Chhaochharia and Grinstein 2009). Of the evidence that does exist of nonprofit

organizations, it is apparent that an absence of strong governance mechanisms may have a

notable and adverse impact on executive pay levels. For example, Brickley, Van Horn, and

Wedig (2010) find that “insider” (non-independent) nonprofit boards are associated with greater

executive compensation, while a comprehensive study of board size by Aggarwal, Evans, and

Nanda (2012) reveals that larger nonprofit boards are associated with weak managerial

incentives. Related work by Cardinaels (2009) further substantiates that a nonprofit board of

poor quality is associated with excessive executive pay.

II.4. The intersection of nonprofit executive pay, firm performance, and governance quality.

Additional and more sophisticated research on the intersection of executive

compensation, organizational performance, and governance quality in nonprofit entities is

warranted for several reasons. First and foremost, extant literature fails to provide a definitive

directional prediction as to how nonprofit CEO pay should relate to organizational performance,

particularly after controlling for governance quality. One reason for this deficiency is simply

that the volume and depth of governance data were lacking prior to the 2008 release of a

redesigned and much more detailed IRS Form 990. By using data originating from the

redesigned Form 990, my study overcomes governance-related data limitations.

A second motivation for further study of the relation between nonprofit pay,

performance, and governance is that different empirical studies utilize different measures of

performance, which further complicates any attempt to make meaningful conclusions on the

nonprofit pay-for-performance link. Uncertainty underlying how to best measure nonprofit

performance – particularly relative to for-profit performance – has long been recognized as

16

problematic (Roomkin and Weisbrod 1999). I overcome these ambiguities by making use of

multiple measures of nonprofit performance.

Lastly, the link between executive pay, performance, and governance has become

increasingly important in response to recent regulatory reforms – most notably, the Dodd-Frank

Act. In the for-profit executive compensation literature, researchers have responded by relating

firm performance to measures of relative CEO pay. Examples include Faleye, Reis, and

Venkateswaran (2013) who utilize a “CEO-employee relative pay [ratio],” defined as the ratio of

CEO pay to average employee pay (in log form), as well as Bebchuk, Cremers, and Peyer (2011)

who study the “CEO pay slice,” computed as the ratio of CEO pay to total pay to the top 5

executives. Nonprofit organizations are keenly aware of the fact that similar disclosure

requirements, such as a ratio of CEO-to-employee relative pay, are likely forthcoming. By using

a definition of relative CEO pay, my paper offers new and timely insights into nonprofit

compensation structures.

III. SAMPLE

Data for this study are based on year 2008 and 2009 filings of IRS Form 990, Return of

Organization Exempt from Income Tax, the annual information return required of most tax-

exempt organizations.4 The form underwent a significant revision for tax year 2008; most

notably, a new section entitled, “Governance, Management, and Disclosure” was added, and

touted as the form’s “crown jewel” by Steven T. Miller, former Commissioner of the Tax

Exempt and Government Entities Division (Miller 2008). Because governance data were

recently added (2008), and are presently available for tax years 2008 and 2009 only, this study’s

4 Private foundations file a Form 990-PF, and small nonprofits (other than private foundations that have gross

receipts less than $100,000 and total assets less than $250,000) file a 990-EZ. For the Form 990 and related tax

documents, see http://www.irs.gov/uac/Form-990,-Return-of-Organization-Exempt-From-Income-Tax- .

17

data are relatively large in the cross-section but relatively short in the time-series. Governance

variables are drawn from tax years 2008 and 2009, while most other variables are based on one

or more of years 2006, 2007, 2008, and 2009.

As alluded to previously, on December 19, 2007, a dramatically redesigned Form 990

was issued, about which Acting Commissioner of the IRS, Kevin Brown, stated the following:

“We need a Form 990 that reflects the way this growing sector operates in the 21st century. The

new 990 aims to give both the IRS and the public an improved window into the way tax-exempt

organizations go about their vital mission” (Internal Revenue Service 2007). In its redesigned

state, the Form 990 spans eleven pages, not including the many (sixteen) supplemental

schedules. Meeting the increased number and complexity of requirements is sure to require

additional time and effort on the part of nonprofits, the disdain for which is perhaps no better

expressed than through the sentiment of one executive director of a large public charity: “If this

is the annual return we have to file, we don’t want to be tax-exempt anymore” (Hopkins 2009, p.

115).5

A large sample of Form 990 returns are provided in the Internal Revenue Service

Statistics of Income (SOI) microdata files, which are publicly and freely available. For this

study, SOI data are retrieved from the National Center for Charitable Statistics (NCCS), which

cleans and distributes nonprofit data provided by the IRS. In the aggregate, the SOI sample

includes over 90 percent of all nonprofit assets and revenues (Yetman and Yetman 2012). 6

5 The first academic publication to utilize the governance data required on the newly redesigned Form 990 is

Yetman and Yetman (2012). They predict and find better financial reporting quality in well-governed nonprofit

organizations, where the key measure of financial reporting quality considered is the accuracy of reported charitable (program) expenses.

6 Sampling rates range from around 1 percent of small-asset classes to 100 percent of large-asset classes in the SOI

data set. Asset classes are ranges of total assets. Each nonprofit will fall into one such category based on its asset

size. Large asset classes typically include organizations having $10 million or more in total assets, while small

asset classes encompass nonprofits having $1 million or less. For an example, see the presentation “SOI Studies,

18

Given that the SOI represents the largest publicly available nonprofit database, it has been used

in multiple empirical studies (e.g., Core, Guay, and Verdi 2006; Keating, Parsons, and Roberts

2008; Yetman and Yetman 2012).7 To capture data provided in the redesigned Form 990, the

sample considered for this study includes all U.S.-based nonprofit organizations in the 2008 and

2009 tax year SOI databases. CEO identity, compensation, and tenure are gathered from the

Guidestar Premium database which, similar to the NCCS, disseminates Form 990 information

into a digitized, machine-readable form. After merging necessary variables from the NCCS and

Guidestar datasets, excluding observations involving either non-positive executive compensation

(Carroll, Hughes, and Luksetich 2005) or CEO turnovers, 12,345 observations remain. It is

important to note that since the 2008 tax year marked the first year in which the redesigned Form

990 was filed, not all 501(c)(3) nonprofits were required to use the new form. Specifically, for

tax year 2008 (2009), only those organizations with total assets of at least $2.5 million ($1.25

million) or gross receipts of at least $1 million ($500,000) were required to use the redesigned

Form 990. As such, to eliminate the possibility of sample selection bias, I restrict the sample to

only those nonprofits required to file the redesigned form since voluntary filers might tend to be

well-governed, resulting in the elimination of 333 observations. Eliminating apparent coding

inconsistencies (as in Core et al. 2006, Aggarwal et al. 2012, Baber et al. 2002, and others),

observations that are classified as “out-of-scope” by the NCCS, and requiring sufficient data to

estimate the main statistical models results in a maximum sample of 7,374 firm-year

observations and 4,429 unique nonprofits. To mitigate the effect of outliers and any remaining

data errors, all continuous variables are winsorized at the 1st and 99

th percentiles.

Data, and Future Plans” by Tom Petska, Director of SOI operations, where “A Typical SOI Sample Design” is described (http://www.taxadmin.org/fta/meet/07rev_est/papers/Petska.pdf).

7 Although frequently used, the accuracy and reliability of Form 990 data have been questioned (e.g., Froelich and

Knoepfle 1996; Froelich, Knoepfle, and Pollak 2000; Krishnan and Yetman 2011). However, literature exists that

serves to guide researchers on how best to clean Form 990 data (e.g., Bowman, Tuckman, and Young 2012;

Roberts 2005; Thornton and Belski 2010). These procedures have been followed in this paper.

19

IV. EMPIRICAL ANALYSIS

Table 1 provides descriptive statistics of variables used in subsequent empirical analyses.

All variables are defined in Appendix A. The mean (median) values of total assets, total

revenue, and number of employees are $206.6 ($56.3) million, $93.5 ($22.9) million, and 896

(311), respectively. The mean (median) CEO compensation is $329,014 ($221,099), and the

mean (median) CEO-to-employee relative pay ratio is 11.33 (8.05). Comparing relative pay

ratios across major industries reveals that the highest values belong to the higher education sector

(mean = 17.69, median = 17.05) and lowest values to the public benefit sector (mean = 5.70,

median = 4.47).8 As for the primary measure of organizational performance utilized in this

paper, the mean (median) proportion of program expenses to total expenses is 84.2% (86.6%).

Comparisons across major industries reveal that the public benefit sector has the highest program

spending ratio (mean = 87.06%, median = 90.21%) while the arts sector has the lowest ratio

(mean = 76.52%, median = 80.07%). In subsequent statistical tests, I account for differences

across industries by including industry fixed effects for each of the 26 major nonprofit industry

groupings (as identified by the National Taxonomy of Exempt Entities (NTEE)), in addition to

industry-adjusting all measures of financial performance.9

The governance index consists of sixteen components assessing several aspects of

nonprofit governance quality, including its governing board, governing policies, compensation

policies, as well as transparency and accountability, and is described in detail in Appendix A.

Most index components consist of “Yes”/ “No” questions coded as 0-1 indicator variables, where

8 “Major industries” at times refer to one of the (larger) 26 NTEE classifications and at other times refer to

aggregated groupings of some of the (smaller) 26 NTEE classifications. See the NTEE-10 Industry

Classifications and NTEE-26 Industry Classifications variable definitions in Appendix A for more details. 9 The industry adjustment to financial performance is computed as the value for firm i in year t less the industry

median in year t, where industries are based on the NTEE’s 26 major groupings.

20

a value of 1 signifies the response hypothesized to indicate better governance quality. Each

component was chosen from the redesigned Form 990 (primarily the “Governance, Management,

and Disclosure” section) on account of: (1) its possible relevance to either CEO compensation or

nonprofit performance based on the for-profit and nonprofit literatures, and (2) the likelihood

that the source question would be answered honestly.10

Questions with especially similar

responses across organizations (specifically, those in which at least 95% of organizations agree

in their responses) are eliminated. Also, in an effort to assign higher weights to components

associated with more diverse responses, I weight each component by its cross-sectional standard

deviation. For each firm-year observation, the governance index is then defined as the ratio of

the sum of the individual components as a proportion of the observation-specific total number of

possible responses, such that governance quality is increasing in the magnitude of the index,

where the index (theoretically) ranges from 0 to 1 in value. The mean (median) standard

deviation weighted governance index is 0.090 (0.089), with the health care sector having the

highest governance index scores (mean = 0.094, median = 0.099) and higher education sector

being associated with the lowest values (mean and median of 0.076).

Table 2 provides correlations among CEO compensation, its possible determinants, and

measures of financial performance and nonprofit governance. As expected, (the log of) the

CEO-to-employee relative pay ratio is positively correlated with both (the log of) the number of

employees and (the log of) total assets, with correlations of 0.69 and 0.45, respectively.

10 For example, the governance index includes an indicator variable for whether at least two-thirds of the

organization’s board of directors is independent. There is a large stream of literature documenting the impact of

board independence on executive pay (e.g., Brickley et al. 2010 for nonprofits; Core et al. 1999 for for-profits). As another example, the governance index also includes indicator variables for whether the nonprofit abides by

strong governing policies related to conflict of interest, whistleblower, and document retention matters. The

importance of having strong governing policies in place has been repeatedly stressed for its potential to curb self-

interested managers’ tendencies (see, for example, the Sarbanes-Oxley Act and recommendations of the Panel of

the Nonprofit Sector).

21

Moreover, the correlation between these two measures of organizational size is also quite high

(0.55), but most other correlations are less than 0.15 in absolute value.

IV.1. Contrasting views of the relationship between nonprofit governance, pay, and performance.

The main focus of the forthcoming empirical analyses is to investigate the intersection of

nonprofit CEO pay, financial performance, and governance quality. Specifically, it may be the

case that CEOs are rewarded for achieving optimal financial performance, whether it be for

maximizing charitable spending (in lieu of administrative or fundraising spending), raising a

significant amount of donations in a cost efficient manner, or ensuring their organization is

productive as evidenced by a large ratio of total revenue per employee. I label this prediction the

Pay-for-Performance Hypothesis. In contrast, in poorly governed nonprofits, CEOs may be free

to prioritize their own best interests rather than those of the organization. These poorly

monitored CEOs may choose to minimize their efforts, resulting in poor performance and/or

paying themselves excessive salaries. I title this prediction the Agency Hypothesis.

To disentangle these predictions about the nature of the relationship between the relative

pay ratio and nonprofit performance, I begin by estimating the following two equations:

Log(Relative Pay) it = β0 + β1 Industry-Adjusted Performance it + β2 Governance Index it

+ ∑ + ∑

+ ∑

+ εit (1)

Industry-Adjusted Performance it = α0 + α1 Log(Relative Pay) it + α2 Governance Index it

+ ∑ + ∑

+ ∑

+ εit (2)

where the industry adjustment is computed as the value for firm i in year t less the industry

median in year t, and where industries are based on the NTEE’s 26 major groupings. Both

models include industry and year fixed effects. Industry-Adjusted Performance is one of four

measures of performance: (1) the ratio of program expenses to total expenses (“Program

22

Spending Ratio”), (2) the ratio of program expenses to total assets (“Program Expenses to

Assets”), (3) the ratio of net fundraising revenue (fundraising revenue net of fundraising

expenses) to total fundraising revenue (“Fundraising Ratio”), and (4) (the log of) the ratio of total

revenue per employee (“Log(Revenue per Employee)”). The Program Spending Ratio has been

utilized as a measure of financial performance in multiple related studies (e.g., Aggarwal et al.

2012; Baber at al. 2002). The ratios Program Expenses to Assets and Fundraising Ratio are

based on performance metrics utilized by Desai and Yetman (2006). Log(Revenue per

Employee) builds upon a related study in the for-profit literature by Faleye et al. (2013).

Detailed definitions of these variables are provided in Appendix A. Subsequently, I consider

several variations of these base models to account for the endogeneity of relative pay and

nonprofit performance.

Regarding equation 1, β1 measures the impact of nonprofit performance on CEO pay,

while β2 captures the effect of governance quality on CEO pay. The Pay-for-Performance

Hypothesis implies a positive β1 coefficient, but offers no directional prediction on the sign of β2.

That is, if all firms are well-governed (i.e., there are no agency problems), then β2 should not

significantly differ from zero since true governance quality does not vary. The Agency

Hypothesis predicts a negative relationship between financial performance and CEO pay, i.e., a

negative β1. The Agency Hypothesis also implies a negative relationship between CEO pay and

governance quality, i.e., a negative β2. Note that these predictions acknowledge the imperfect

nature of the governance index (arising due to non-quantifiable aspects of governance quality); if

instead the governance index measured nonprofit governance quality perfectly and sufficiently,

the Agency Hypothesis would imply a β1 coefficient that is approximately zero and a β2

coefficient that is negative in value, as the governance index would capture the full effect of

23

governance on pay.11

In summary, and in the context of equation 1, the Pay-for-Performance

(Agency) Hypothesis implies a β1 that is positive (negative) and a β2 that is approximately zero

(negative).

Regarding equation 2, α1 measures the impact of CEO pay on nonprofit performance

while α2 captures the effect of governance quality on nonprofit performance. Similar to equation

1, the Pay-for-Performance Hypothesis implies a positive α1 and an α2 that does not significantly

differ from zero. The Agency Hypothesis predicts a negative (positive) relationship between

CEO pay (governance quality) and financial performance, i.e., a negative α1 and positive α2. As

discussed in the context of equation 1, the governance index is presumed to measure true

governance quality imperfectly. If this were not the case, the Agency Hypothesis would imply

an α1 coefficient that does not significantly differ from zero. However, realistically speaking, the

governance index is imperfect and, given the Agency Hypothesis’s expectation of higher pay in

poorly governed nonprofits, the prediction becomes a negative α1. In summary, and with regard

to equation 2, the Pay-for-Performance (Agency) Hypothesis implies an α1 that is positive

(negative) and an α2 that is approximately zero (positive).

IV.2. Determinants of CEO-to-employee relative pay and nonprofit performance.

Table 3 reports the results of pooled regressions where (the log of) the CEO-to-employee

relative pay ratio is regressed on its traditional determinants, nonprofit performance, and

governance quality, in addition to industry and year fixed effects. The table includes separate

regressions for each of the four different measures of financial performance considered. All

11 An alternative prediction about the relationship between CEO pay and governance quality stems from an

empirical finding in the for-profit literature. The argument states that intense monitoring is associated with lower

executive pay levels since shareholders will see less of a need to compensate a CEO who is already held

accountable by way of the company’s strong governance mechanisms (Acharya and Volpin 2010). These

employment conditions would be viewed negatively by the CEO, and could encourage him to bargain for higher

compensation, implying a potentially positive link between CEO pay and governance quality.

24

measures of financial performance are presented in industry-adjusted form. Of primary

importance among Table 3 results is a negative relation between all four measures of financial

performance and the CEO-to-employee relative pay ratio (p-values < 0.01). These preliminary

results lend support to a negative relation between CEO pay and nonprofit performance (and, by

extension, the Agency Hypothesis). According to these results, poor performing nonprofits are

also characterized by excessive pay, implying weak governance and ineffective monitoring

mechanisms.

Table 3 identifies several other important determinants of CEO pay. For example, the

relative pay ratio is positively related to (the log of) the number of employees, (the log of) total

assets, and (the log of) CEO tenure, and is negatively related to government revenue, (the log of)

median state-level household income, an indicator variable set to 1 if the CEO is a female, and

an indicator variable set to 1 for commercial nonprofits (defined as those with at least 90% of

revenues drawn from program services, as in Aggarwal et al. 2012 and Steinberg 2004). The

positive association between CEO pay and organizational size supports a widely documented

finding in the related nonprofit literature (e.g., Aggarwal et al. 2012; Frumkin and Keating 2010;

Hallock 2002). The result that CEO pay is increasing in CEO tenure suggests that entrenched,

long-standing CEOs may be able to extract higher compensation or, alternatively, could

represent additional pay for experience. One possible explanation for the negative association

between relative pay and government revenue is that greater reliance on the government for

resources corresponds to closer monitoring, and thus more reasonable CEO compensation levels.

Average employee pay is positively associated with median state household income but, by

construction, inversely related to the relative pay ratio, thus the negative association between the

two. The (female) gender indicator variable enters with a negative and significant coefficient.

25

One possible reason for this result is that women are more willing to accept lower paying

positions when operating in a charitable environment. The negative and significant coefficient

on the commercial nonprofit indicator variable is likely due to differences between the

commercial and traditional sub-groups on several dimensions, such as organizational size.

Table 3 also provides evidence that good governance quality may alleviate excessive

executive pay, with a statistically significant association documented in columns I and II (p-

values < 0.05). For example, based on column I results, a single standard deviation improvement

in the governance index (0.024) implies an approximate 1.4% reduction in the relative pay

ratio.12

Additionally, the results indicate a non-linear relation between the relative pay ratio and

board size, suggesting that CEOs of nonprofits having larger boards are associated with higher

relative pay ratios. The economic significance of the relation between governance quality and

the relative pay ratio will be explored in more detail in subsequent analyses. All other variables

either enter insignificantly or alternate signs through the four model specifications.

The rather lack-luster results with respect to the relationship between governance quality

and CEO compensation may come as a surprise. However, most of the signs and significance of

coefficients in Table 3 support the Agency Hypothesis, particularly the highly negative and

significant association between CEO pay and nonprofit performance. These results clearly

indicate agency (i.e., governance) problems in many nonprofits; that is, the relatively weak

results for the governance index are likely due to its inability to measure governance quality

perfectly. As such, it may be the case that the IRS’s approach to measuring nonprofit

governance quality is misleading, or simply that nonprofit entities answer the questions

12 Given the (unscaled) coefficient on the Governance Index of –0.590, a one standard deviation increase in the

Governance Index (0.024) implies a e(–0.590 × 0.024) – 1 ≈ –1.4 % change in the relative pay ratio.

26

dishonestly. These possibilities raise some concern about the efficacy of a new and costly set of

disclosure requirements.

Table 4 provides the results of pooled regressions where four different measures of

(industry-adjusted) nonprofit performance are regressed on traditional determinants, (the log of)

the CEO-to-employee relative pay ratio, and measures of governance quality, in addition to

industry and year fixed effects. Similar to Table 3, this table includes a separate column for each

of the four different measures of financial performance.

The most relevant finding arising from Table 4 results is a negative relationship between

the relative pay ratio and all four measures of financial performance (p-values < 0.01). In terms

of economic significance, a one standard deviation increase in (the log of) the relative pay ratio

(0.846) implies a 0.8% decrease in the (industry-adjusted) Program Spending Ratio.13

Comparable calculations for the remaining three performance metrics reveal that a one standard

deviation increase in the relative pay ratio implies reductions of 6.7%, 0.9%, and 17.4% in the

(industry-adjusted) Program Expenses to Assets ratio, Fundraising Ratio, and Log(Revenue per

Employee).14

These results, though still preliminary, lend further support to a negative relation

between relative pay and nonprofit performance in favor of the Agency Hypothesis.15

IV.3. Endogeneity of CEO pay and nonprofit performance.

13 Given the (unscaled) coefficient of –0.010 on Log(Relative Pay) in column I of Table 4 (where the dependent

variable is the Program Spending Ratio), a one standard deviation increase in Log(Relative Pay) implies a –0.010

× 0.846 ≈ –0.8% change in the Program Spending Ratio. 14 Given (unscaled) coefficients of –0.079, –0.011, and –0.206 on Log(Relative Pay) in columns II, III, and IV of

Table 4, a one standard deviation increase in Log(Relative Pay) implies –0.079 × 0.846 ≈ –6.7%, –0.011 × 0.846 ≈

–0.9%, and –0.206 × 0.846 ≈ –17.4% changes in Program Expenses to Assets, Fundraising Ratio, and

Log(Revenue per Employee), respectively. 15 Although equations 1 and 2 have several explanatory variables in common, there are control variables exclusive to

equation 1 (i.e., Log(Household Income), Log(CEO Tenure), and Gender Dummy) and equation 2 (i.e.,

Fundraising Dummy and Total Liabilities/Total Assets). Therefore, although the results associated with each of

equations 1 and 2 contribute new information, I acknowledge the fact that the two estimations are not

independent.

27

As implied by the model specifications in equations 1 and 2, both relative pay and

financial performance are likely endogenous. Indeed, according to the Agency Hypothesis, both

are impacted by governance quality and, according to the Pay-for-Performance Hypothesis, both

are impacted by CEO skill or effort. I address this possibility using three empirical methods: (1)

two-stage least squares procedures where financial performance and, separately, the relative pay

ratio are treated as endogenous; (2) a changes specification, where changes are computed as the

difference between time t and time t–1 values; and (3) modeling the relative pay ratio as a

function of lagged (rather than contemporaneous) financial performance. While these methods

represent a targeted attempt at addressing concerns about endogenous associations, I do not

claim to eliminate endogeneity altogether. However, it is worth noting that the two endogenous

variables of interest, the relative pay ratio and nonprofit financial performance, consistently

exhibit a significant negative relationship. Thus, any latent variables that would otherwise lead

to a positive and potentially spurious association do not seem to be of great concern, given that I

find the two variables to be negatively related.

The first approach I use to address endogeneity is to estimate the relations between CEO

pay and firm performance via a two-stage procedure. Treating nonprofit financial performance

as endogenous, I adopt two instrumental variables: the Total Liabilities/Total Assets ratio lagged

by one year and financial performance lagged by two years.16

The ratio of total liabilities to total

assets is a measure of financial reporting quality (e.g., Krishnan, Yetman, and Yetman 2006;

Keating, Parsons, and Roberts 2008), and is expected to affect the relative pay ratio only through

16

There are two exceptions to these instruments. First, when instrumenting for the financial performance measure

Program Expenses to Assets, I use the prior-period value of the industry median Total Liabilities/Total Assets ratio. Using the industry median avoids a mechanical relationship between interchanging two firm-specific

variables having shared denominators and, by extension, reduces over-identification potential. Second, the data

do not permit calculating the twice-lagged value of Log(Revenue per Employee); specifically, there are no

observations with data for the number of employees at time t–2. As a result, as instruments for Log(Revenue per

Employee), I use the twice-lagged values of the firm-specific and industry median Log(Total Revenue).

28

financial performance, thus lending to its validity as an instrument. Table 5 presents two-stage

least squares (2SLS) regression results where nonprofit performance is treated as endogenous.

The Shea’s partial R2’s of the two instruments in the first stage regression provide evidence that

they are jointly correlated with financial performance (partial R2’s ranging from 24.5% to

73.3%). Moreover, the Sargan statistic testing for the over-identification of both instruments

fails to reject the null hypothesis that the instruments as a group are exogenous (p-values of

0.128 or greater). After addressing endogeneity concerns, Table 5 results uphold the previously-

documented negative relation between the CEO-to-employee relative pay ratio and financial

performance, with all four coefficients negative in sign and statistically significant at the 0.05

level.

Table 6 presents the results for when the relative pay ratio is treated as endogenous. I

adopt two variables to instrument for relative pay: the industry median value of (the log of) the

ratio of CEO compensation to the number of volunteers, and the twice-lagged value of (the log

of) the ratio of CEO compensation to total compensation paid to all employees of the

organization.17

The Shea’s partial R2’s of the two instruments in the first stage regression

demonstrate that they are jointly correlated with the relative pay ratio (partial R2’s ranging from

40.1% to 43.1%). Additionally, the Sargan statistic testing for the over-identification of both

instruments fails to reject the hypothesis that the instruments as a group are exogenous (p-values

of 0.270 or greater). After taking into account the endogenous nature of the relative pay ratio,

Table 6 results support the previously-documented negative relation between financial

17 Insufficient data on the number of employees at time t–2 prohibits computing the twice-lagged value of the

relative pay ratio (the number of employees is needed to compute the denominator of the pay ratio). Therefore, in

place of the twice-lagged value of the relative pay ratio, I use the twice-lagged value of (the log of) the ratio of

CEO compensation to total compensation.

29

performance and the relative pay ratio, with all coefficients negative in sign and highly

statistically significant for all four measures of financial performance (p-values < 0.01).

For the remainder of the paper, I focus on a single measure of performance: the program

spending ratio. One reason to prioritize this performance metric over competing alternatives is

for its conceptually appealing qualities. Specifically, nonprofit organizations are expected to

fulfill a charitable mission, the core expenses for which are captured in program expense

categories. Thus, it is reasonable to conclude that a nonprofit devoting a high proportion of its

total spending towards charitable causes (relative to administrative or fundraising alternatives) is

performing well. Another reason to focus on the program spending ratio is a matter of

precedence, given that it is a widely-accepted financial performance metric utilized in related

literature (e.g., Aggarwal et al. 2012; Baber et al. 2002).

As an additional check of endogeneity between the relative pay ratio and the program

spending ratio, I estimate equation 2 in change form. Results, provided in Panel A of Table 7,

reveal a significant negative relation between the change in relative pay and the change in

program spending (p-value = 0.007). Other results of significance include a significant positive

relation between changes in program spending and in (the log of) the nonprofit’s age, indicating

that organizations devote more resources toward charitable activities as they age, particularly

relative to alternative uses of funds (i.e., administrative or fundraising). The significant negative

association between changes in program spending and in liquid assets suggests that nonprofit

managers do not increase their charitable spending levels in response to increases in liquidity.

As a further test of endogeneity, I estimate equation 2 using the lagged relative pay ratio

in place of its contemporaneous value. Results, offered in Panel B of Table 7, demonstrate a

negative pay-for-performance link; that is, managers having high relative pay ratios tend to

30

spend relatively little on charitable activities (p-value = 0.058). All other variables enter with

coefficients and statistical significance levels that are similar to those reported in column I of

Table 4, lending further support to a negative pay-for-performance link.

IV.4. Executive perquisites, nonprofit performance, and governance.

The results presented in this paper support a negative relationship between CEO pay and

financial performance. How do other forms of managerial incentives relate to nonprofit

performance? In this section, I analyze the relationship between executive perquisites, financial

performance, and governance quality. One aspect of the revised Form 990 is more detailed

information on executive compensation. For example, the first question on Schedule J

(“Compensation Information”) asks whether the organization provided perquisites to any officer,

director, trustee, key employee, or highly compensated employee. Among the eight listed

perquisite categories are first-class or charter travel, travel for companions, a discretionary

spending account, and housing allowances, to name a few.

To begin, I estimate a multivariate logit model predicting the probability of at least one

type of perquisite being awarded to an executive in a given year. Subsequently, I consider the

probability that at least one executive is awarded first-class or charter travel – a specific and

indisputably luxurious type of perquisite that is conceivably highly sensitive to governance

quality. In both models, the independent variables of interest include (1) an indicator variable

for superior nonprofit performance, where performance is measured as the program spending

ratio, (2) governance quality, captured by the governance index, and (3) the interaction of the

variables described in (1) and (2). The models control for traditional determinants of the relative

pay ratio, where determinants are those referred to in equation 1 with baseline results in column I

of Table 3.

31

Column I of Table 8 presents the results of the logit regression predicting the probability

that at least one type of perquisite is awarded, while column II exclusively focuses on the

perquisite type of first-class or charter travel. The direction and significance of the coefficients

of interest are highly comparable between the two models. The indicator variable capturing

superior nonprofit performance enters with a negative and highly statistically significant

coefficient (p-value < 0.0001). This result suggests that superior financial performance (and

effective monitoring mechanisms) is inversely related to the likelihood of receiving executive

perquisites, which supports the earlier result of a negative relation between nonprofit

performance and reportable compensation. The governance index also enters with a negative

and significant coefficient (p-value < 0.0001), indicating that well-governed, effectively

monitored firms are less apt to award perquisites to managing executives. Of particular

importance is the interaction term between these two variables, which enters with a positive and

highly statistically significant coefficient (p-value < 0.0001). This latter result suggests that,

although high-performing and well-governed firms are individually less likely to award

executive perquisites, good governance quality is associated with an incremental but significant

increase in the likelihood that executives of high-performing nonprofits will be rewarded with

perquisites (relative to their poorly-governed peers).

I also investigate the relationship between the variety of executive perquisites, nonprofit

performance, and governance quality. Specifically, I estimate a multivariate ordered probit

model based on the number of types of perquisites identified on the Form 990 as being awarded

during the year. Since the Form 990 lists eight perquisite categories, the dependent variable

ranges in value from 0 to 8, in increments of 1. All independent variables are the same as those

utilized in the logit model of executive perquisites. Results, reported in column III of Table 8,

32

include coefficients that enter with signs and statistical significance levels that are very similar to

those corresponding to the logit models. Specifically, I find that the variety of executive

perquisites decreases in cases of superior nonprofit performance and in cases of good governance

quality, but that a well-governed nonprofit is associated with an incremental but significant

increase in a high-performing nonprofit organization’s variety of executive perquisites (relative

to their poorly-governed counterparts). Collectively, the results in Table 8 lend further support

to a negative association between managerial incentives and nonprofit performance that is

incrementally impacted by the organization’s governance quality.

IV.5. CEO pay and nonprofit performance surrounding CEO turnovers.

The evidence presented so far in this paper demonstrates a negative relation between

nonprofit CEO pay and financial performance, consistent with the Agency Hypothesis. A

relevant question then becomes, what are the observed patterns in CEO pay and nonprofit

performance surrounding CEO turnovers? I isolate all instances of CEO turnovers, both forced

and voluntary.18

I match each turnover (“treatment”) firm to a comparable non-turnover

(“control”) firm, where the control firm is required to have been in the same industry and of

similar size (total assets) in the year prior to the treatment firm’s turnover. To qualify as a

match, I allow a maximum 10% differential between treatment and control firms’ total assets.

Panel A of Table 9 presents univariate analyses comparing the industry-adjusted program

spending ratios of turnover firms to those of non-turnover firms. A review of the “Difference”

column reveals that turnover firms experienced significantly lower financial performance in the

year prior to (p-value = 0.006) and the year of (p-value = 0.001) turnover relative to the sample

18 Due to data limitations, I am not able to decipher whether a CEO turnover was forced or voluntary.

33

of control firms. Interestingly, this disparity does not resolve following turnover; that is,

turnover firms still report significantly lower performance values relative to matched firms.

Panel B of Table 9 reports the results of multivariate logit regressions, where the

probability of turnover is modeled as a function of prior-period nonprofit financial performance

and the relative pay ratio. Specifically, to capture the effect of prior-period performance on the

likelihood of CEO turnover, I define an indicator variable that is set to 1 when the nonprofit

exceeded the annual industry median value of the program spending ratio, and is 0 otherwise. I

then include an interaction term between the high performance indicator and relative pay as a

means to capture the incremental effect of superior financial performance on the relation

between relative pay and the probability of turnover. The logit model also includes control

variables for firm- and CEO-specific attributes, in addition to industry fixed effects. The results

provided in Panel B of Table 9 indicate a notably higher probability of turnover when relative

pay ratios are large (p-value = 0.041). Interestingly, the interaction term (between relative pay

and the high performance indicator) enters with a negative and statistically significant coefficient

(p-value = 0.067) suggesting that relative to their poor-performing counterparts, highly-paid

CEOs of high-performing nonprofits are associated with an incrementally lower probability of

turnover. The results presented in Table 9 suggest that, although nonprofits do not necessarily

pay for performance, achieving optimal financial performance is associated with a lesser chance

of CEO turnover.

IV.6. The role of governance in offsetting the consequences of agency conflicts.

An important result arising from earlier empirical analyses is that higher CEO pay levels

are associated with inferior corporate governance quality. In particular, columns I and II of

Table 3 reveal a statistically significant inverse relation between the governance index and

34

relative pay (p-values of 0.040 and 0.048 for the Program Spending Ratio and Program Expenses

to Assets models, respectively). I next examine the economic significance of this result.

Superior corporate governance should prove most valuable when agency conflicts are

high. That is, in the context of this paper, it is reasonable to expect governance mechanisms to

play a more significant role in mitigating the excessive compensation of CEOs when the

organization is marred by agency problems. To test this prediction, I classify a nonprofit as

being characterized by agency problems when its level of liquid assets is excessively high. In a

study of the agency problems of excess liquid assets, Core, Guay, and Verdi (2006) model

nonprofits’ liquid asset levels as a function of firm size, revenue volatility, and access to debt.19

As Core et al. describe, there are at least three possible reasons underlying an abnormally high

level of liquid assets: (1) the nonprofit anticipates needing to finance future growth

opportunities; (2) the nonprofit benefits from a strong monitoring role; or (3) the nonprofit is

tainted by agency problems. The authors’ evidence strongly favors the third reason (i.e., that

excess liquid assets are indicative of agency problems), which aligns well with a large strand of

literature in the for-profit sector (e.g., Harford 1999; Opler, Pinkowitz, Stulz, and Williamson

1999). As such, following the empirical model employed by Core et al., I compute a nonprofit’s

excess liquid assets as the residual from the following model:

Liquid Assets/Total Expenses it = β0 + β1 Log(Total Revenue) it + β2 CV(Total Revenue) it+ β3 Access to Debt it

+ β4 CV(Total Revenue) it Access to Debt it

+ ∑

+ ∑

+ εit (3)

where CV(Total Revenue) is the coefficient of variation of total revenue, defined as the ratio of

the standard deviation of total revenue to mean total revenue measured over the five-year period

ending in year t, and Access to Debt is an indicator variable coded as 1 if the firm utilized debt

19 The definition of “liquid assets” employed in this paper is termed “endowment assets” in Core et al. (2006).

35

(tax-exempt bonds) during the ten-year period ending in year t. Equation 3 is estimated yearly,

and the continuous variables (Log(Total Revenue) and CV(Total Revenue)) are winsorized at the

1st and 99

th percentiles to help mitigate the impact of outliers. I require a minimum of four years

to compute CV(Total Revenue) and rely on the Guidestar Premium database for the relatively

longer time span of data needed to estimate this equation.20

Of primary importance is gaining an understanding of the impact of governance on the

CEO pay levels of nonprofit organizations with potentially serious agency problems. Using

estimates of excess liquid asset levels, I capture probable agency conflicts through an indicator

variable, Agency Dummy, defined as 1 (0) for the upper (lower) quartile of excess liquid assets. I

then estimate a variant of equation 1 that includes the Agency Dummy, components of the

governance index in place of its aggregate form, and interaction terms between the Agency

Dummy and each governance mechanism. Results, reported in Table 10, reveal several

associations between measures of governance quality and the relative pay ratio that significantly

vary according to the potential severity of agency conflicts.

In the final portion of this analysis, I compute the economic significance of governance

variables whose statistical associations with the relative pay ratio are significantly related to the

potential severity of agency conflicts. Specifically, from column II of Table 10, these

governance variables include Board Independence, Conflict of Interest Policy, Whistleblower

20 For reference, Table B9 of Appendix B presents pooled regression results of the liquid asset expectations model

specified in equation 3. The pooled model is presented for the full sample for which sufficient data are available

during the 2008-2009 period. Year fixed effects are incorporated in the pooled model but are excluded when

estimating equation 3 on a yearly basis to obtain residual (excess) liquid asset levels. Coefficient estimates are

generally comparable to those reported by Core et al. (2006). In sum, as revenue (and cash flow) becomes more

volatile, liquid asset levels increase. In contrast, lower liquid asset levels are found in larger nonprofits.

Interestingly (and contrary to Core et al.), access to debt is associated with higher debt levels for my sample entities, suggesting that nonprofits issue debt in an effort to build up liquid reserves. The negative and statistically

significant coefficient on the interaction term, CV(Total Revenue) Access to Debt, indicates that having access

to external debt markets offsets a nonprofit’s tendency to build up liquid assets in response to a revenue stream

that is particularly volatile.

36

Policy, CEO Compensation Policy, and Non-CEO Compensation Policy.21

Column I (III) of

Table 11 reports the marginal effects of these five governance characteristics for nonprofits in

the lower (upper) quartile of excess liquid asset levels, denoted as the “Low Agency group”

(“High Agency group”). Column II provides the incremental difference between the two groups.

Estimated differences in the CEO-to-employee relative pay ratio appear economically significant

across the Low Agency and High Agency groups. For example, with respect to governing

policies, the decision to adopt a conflict of interest policy is associated with an impressive 15.4%

decrease in the relative pay ratio when liquid asset levels are high, but implies an increase of

13.8% when excess liquid asset levels are low.22

Relatedly, for the High Agency (Low Agency)

group, the decision to adopt a whistleblower policy is associated with a 14.1% decrease (3.6%

increase) in relative pay, while the decision to adopt policies pertaining to CEO compensation

implies a 4.0% decrease (19.5% increase) in the relative pay ratio. These results are to be

expected, given that both the Sarbanes-Oxley Act and the Panel on the Nonprofit Sector (a

coalition of leaders from U.S. charitable organizations) cited agency problems as a key

motivating factor when advocating for the adoption of strong governance policies. Regarding

board structure, a switch in response from “No” to “Yes” as to whether at least two-thirds of the

nonprofit’s board members are independent is associated with a decrease in relative pay of 1.4%

for the High Agency group, but suggests a 35.9% increase for the Low Agency group. This

inverse and economically meaningful relationship between independent board members and

executive pay among entities characterized by potentially severe agency conflicts is consistent

21

Due to a relatively low response rate, there are two components of the governance index that are omitted from the

analysis: Executive Reimbursement Policy and Executive Substantiation Policy, each having only 2,459 responses (compared to 4,000+ responses for other index components).

22 The economic impact of switching from “No” to “Yes” on the Conflict of Interest Policy question for “Low

Agency” nonprofits = e0.1288 – 1 = 13.75%, where 0.1288 is the (unscaled) coefficient on the Board Independence

variable (see column I of Table 10). For details on how other measures of economic significance were computed,

see the description to Table 11.

37

with related literature of the for-profit (e.g., Chhaochharia and Grinstein 2009) and nonprofit

(e.g., Brickley et al. 2010) sectors. Taken as a whole, the results in Table 11 demonstrate the

ability of several governance mechanisms to alleviate excessive executive pay among nonprofits

marred by potentially severe agency conflicts.

IV.7. The impact of governance on financial performance surrounding changes in relative pay.

The empirical results in this paper suggest that CEO pay is negatively related to nonprofit

performance, yet good governance quality helps to offset excessively high relative pay ratios.

Next, I examine the relation between changes in nonprofit financial performance with regard to

changes in the relative pay ratio for different levels of governance quality. That is, I analyze

whether governance quality counteracts the negative relation between relative pay and financial

performance. Table 12 reports the results. Although changes in relative pay and changes in

program spending are statistically significantly negatively related, I find that – relative to their

poorly-governed peers – well-governed nonprofits enjoy an incremental improvement in

financial performance in response to changes in the relative pay ratio (p-value = 0.047). This

finding lends further credibility to the notion that adopting stronger governance mechanisms

translates into more favorable corporate outcomes.

V. ROBUSTNESS CHECKS

I next provide several empirical tests to validate the robustness of the results.

Specifically, I re-estimate equations 1 and 2 under the following settings: (1) after splitting the

sample into commercial and traditional nonprofits; (2) using alternative measures of

organizational size; (3) using alternative measures of CEO-to-employee relative pay; and (4)

38

after including state fixed effects. Tabulated results are presented in Appendix B, with a

discussion of the motivations and results of each robustness check presented below.

V.1. Commercial versus traditional nonprofit organizations.

In the first of a series of robustness checks, I re-estimate equations 1 and 2 for sub-

samples of “commercial” and “traditional” nonprofit organizations. “Commercial” nonprofits

are defined as those with at least 90% of revenues from program services (e.g., Aggarwal et al.

2012; Steinberg 2004). In contrast, “traditional” nonprofits operate on a greater variety of

revenue sources such as donations, government grants, and rental income. The commercial

versus traditional nonprofit classification tends to cluster along industry lines. For example, the

hospital industry is dominated by commercial nonprofits (88%), while traditional nonprofits

dominate the arts and public benefit industries (97% and 89%, respectively). On average,

commercial nonprofits tend to be larger compared to traditional nonprofits in terms of the

number of employees (mean for commercial = 1,454; mean for traditional = 557) and total

revenue (mean for commercial = $152.6 million; mean for traditional = $57.6 million), but

comparable in terms of total assets (mean for commercial = $206.9 million; mean for traditional

= $206.4 million). As for compensation, CEOs of commercial nonprofits benefit from a higher

average salary (mean = $453,895; median = $307,419) relative to the average pay for CEOs of

traditional nonprofits (mean = $253,138; median = $185,530).

Given their extreme reliance on a single source of income, it is reasonable to expect

commercial nonprofits to attempt to be better stewards of their resources, including CEO

compensation-related matters. One might also expect that commercial nonprofits operate in a

more competitive environment, thus leading to potentially better financial performance

outcomes. Moreover, both differences could plausibly affect the nature of the relationship

39

between the relative pay ratio and financial performance. Earlier regression specifications

incorporated an indicator variable, Commercial Dummy, to control for these potential

differences. As another way to check whether the negative pay-for-performance relation holds

for commercial versus traditional nonprofits, I re-estimate equations 1 and 2 after splitting the

sample into commercial and traditional sub-groups. Results, provided in Tables B1 (equation 1)

and B2 (equation 2) of Appendix B, demonstrate that the statistically significant negative

association between the relative pay ratio and all four measures of financial performance applies

to commercial and traditional nonprofits alike.23

V.2. Alternative measures of organizational size.

Existing studies on the compensation of nonprofit executives uniformly agree on

organizational size as being a critically important determinant of pay levels (e.g., Aggarwal et al.

2012; Frumkin and Keating 2010; Hallock 2002). The baseline models utilized in this paper

depend upon two variables to control for nonprofit size: Log(No. of Employees) and Log(Total

Assets). I re-estimate equations 1 and 2 using two sets of alternative measures of size: (1) the

number of employees and its squared term, along with total assets and its squared term; and (2)

the number of employees and its reciprocal, along with total assets and its reciprocal. Tables B3

(equation 1) and B4 (equation 2) present the results, which verify a negative link between the

relative pay ratio and all four measures of financial performance, regardless of what empirical

proxies are used to control for size. Coefficients on the size terms are also informative.

Specifically, results reported in columns I, III, V, and VII of Table B3 demonstrate that relative

pay is indeed positively associated with nonprofit size, but that this effect diminishes for

particularly large entities. A similar pattern is observed when considering nonprofit financial

23 The only exception is in column (VI) of Table B2, where the association between the Fundraising Ratio and

Log(Relative Pay) is still negative, but is not statistically significant at conventional levels (p-value = 0.179).

40

performance (see columns I, III, V, and VII of Table B4), with the exception of a mechanical,

negative relation between total assets and the performance measure Program Expenses to Assets.

In sum, the negative association between the CEO-to-employee relative pay ratio and financial

performance stands independent of the variables chosen to control for nonprofit size.

V.3. Alternative measures of CEO-to-employee relative pay.

The inferences of this paper rely on a specific measure of CEO compensation: the CEO-

to-employee relative pay ratio. In this robustness test, I investigate whether the negative relation

between pay and performance holds under settings of different (though related) measures of

relative pay. Specifically, I re-estimate equation 2 using alternative measures of CEO pay,

including the industry-adjusted (log of the) CEO-to-employee relative pay ratio and the excess

form of the relative pay ratio, where the latter is defined as the residual from a variant of

equation 1 (see the definition of Excess Compensation in Appendix A). Table B5 presents the

results corresponding to the first alternative definition of CEO compensation and Table B6

reports the results corresponding to the latter definition. Using either alternative measure of pay,

a statistically significant negative relation between relative pay and all four measures of

nonprofit financial performance is noted.

V.4. Inclusion of state fixed effects.

Various states across the U.S. have mandated particular governance policies of their local

nonprofit organizations (e.g., Desai and Yetman 2006). The premise of this paper is to

investigate how the voluntary adoption of governance mechanisms affects both the relative pay

ratio and financial performance, as well as the relation between the two. Therefore, to address

the mandatory nature of some governance characteristics by U.S. states, I re-estimate both

equations 1 and 2 after incorporating state fixed effects. Results are reported in Tables B7

41

(equation 1) and B8 (equation 2). These outcomes are qualitatively very similar to earlier

findings, including a negative pay-for-performance relation and the ability of good governance

quality to offset excessive CEO compensation.

VI. CONCLUSION

Allegations of grossly overpaid nonprofit CEOs have become increasingly common, yet

whether this excessive pay is typically established in an optimal or sub-optimal manner remains

an unanswered question. The efforts of a capable CEO who prioritizes his organization’s

responsibility to serve the community should be easily recognizable through superior nonprofit

performance and, by extension, through greater compensation to the CEO. In contrast, a

nonprofit marred by agency conflicts whereby the CEO prioritizes his own interests before those

of his organization is likely to suffer poor financial performance, even though the CEO will tend

to extract excessive rents on account of his power and influence. In an effort to disentangle these

predictions, I investigate whether excessive CEO pay is associated with superior or inferior

nonprofit financial performance. Moreover, I examine this relation in the context of several

measures of nonprofit financial performance, as well as in terms of a timely and relevant

definition of CEO compensation which is measured relative to the average employee’s salary.

That is, relative pay measures are becoming increasingly relevant given recent regulation (e.g.,

the Dodd-Frank Act) and their proposed reporting requirements.

I document a statistically significant and economically meaningful relation between

nonprofit CEO pay and financial performance. Modeling financial performance as a function of

the CEO-to-employee relative pay ratio and its many determinants – including those that

potentially impact either CEO compensation or financial performance, I find that a one standard

42

deviation in the relative pay ratio is associated with notably lower performance in the order of

approximately 1%, 7%, 1%, and 17% when considering the (industry-adjusted) ratios of program

expenses to total expenses, program expenses to total assets, net fundraising revenue to total

fundraising revenue, and (the log of) total revenue per employee, respectively. These results are

robust to a number of statistical tests that address the endogenous nature of the relationship

between the relative pay ratio and nonprofit financial performance, including two-stage least

squares, a changes specification, and in modeling performance as a function of the lagged (rather

than contemporaneous) relative pay ratio. The finding of a negative association between the

relative pay ratio and nonprofit performance is suggestive of agency conflicts.

I explore this negative pay-for-performance relation further by considering other settings

and related predictions. First, I examine whether the negative pay-for-performance relation

holds for executive perquisites (rather than reportable compensation). The results indicate that a

higher relative pay ratio is accompanied by a lower likelihood that executives will be awarded

perquisites, though good governance quality increases this probability when the nonprofit is also

performing well. Next, I investigate pay and performance implications surrounding CEO

turnovers. I find a higher probability of turnover when the CEO-to-employee relative pay ratio is

excessively high, though this probability is significantly lower when the CEO achieved superior

financial performance in the year preceding his departure. Additionally, I gauge the importance

of governance quality in mitigating excessive executive pay, with a focus on nonprofits

suspected of suffering severe agency problems. The results point to several governance factors

offering an economically meaningful impact on the relative pay ratio. In particular, among

nonprofits with potentially severe agency conflicts, adopting a conflict of interest policy

(whistleblower policy) is associated with a reduction in the relative pay ratio in the order of 15%

43

(14%). Additionally, although changes in the relative pay ratio and changes in charitable

spending are negatively related, I find that a nonprofit that is well-governed realizes an

incremental but statistically significant increase in performance pursuant to changes in the

CEO’s compensation contract. Collectively, these findings support a negative association

between the CEO-to-employee relative pay ratio and nonprofit performance that differs on

account of a nonprofit’s governance quality and, relatedly, its extent of firm-specific agency

conflicts.

The results in this paper demonstrate a negative relationship between the CEO-to-

employee relative pay ratio and nonprofit financial performance, suggestive of agency conflicts.

Fortunately, nonprofits with severe agency problems can mitigate excessive executive pay by

adopting strong corporate governance mechanisms. Thus, while this paper validates concerns

that some nonprofit CEOs are able to extract more compensation than they deserve, it also sheds

light on the importance of corporate governance, particularly since good governance quality has

the potential to alleviate the undesirable consequences of firm-specific agency conflicts.

44

REFERENCES

Aggarwal, R. K., Evans, M. E., and D. Nanda. (2012). Nonprofit Boards: Size, Performance and

Managerial Incentives. Journal of Accounting and Economics 53(1-2): 466-487.

Alexander, J. A., and S. D. Lee. (2006). Does Governance Matter? Board Configuration and

Performance in Not-for-Profit Hospitals. The Milbank Quarterly 84(4): 733-758.

American Bureau of Shipping. (2011). Internal Revenue Service Form 990, Return of

Organization Exempt from Income Tax. Retrieved from

http://990s.foundationcenter.org/990_pdf_archive/134/134921556/134921556_201112_9

90O.pdf

Baber, W. R., Daniel, P. L., and A. A. Roberts. (2002). Compensation to Managers of Charitable

Organizations: An Empirical Study of the Role of Accounting Measures of Program

Activities. The Accounting Review 77(3): 679-693.

Ballou, J. P., and B. A. Weisbrod. (2003). Managerial Rewards and the Behavior of For-Profit,

Governmental, and Nonprofit Organizations: Evidence from the Hospital Industry.

Journal of Public Economics 87(9-10): 1895-1920.

Bebchuk, L. A., Cremers, K. J. M., and U. C. Peyer. (2011). The CEO Pay Slice. Journal of

Financial Economics 102(1): 199-221.

Bertrand, M., and A. Schoar. (2003). Managing with Style: The Effect of Managers on Firm

Policies. The Quarterly Journal of Economics 118(4): 1169-1208.

Bowman, W., Tuckman, H. P., and D. R. Young. (2011). Issues in Nonprofit Finance Research:

Surplus, Endowment, and Endowment Portfolios. Nonprofit and Voluntary Sector

Quarterly 41(4): 560-579.

Brickley, J. A., and R. L. Van Horn. (2002). Managerial Incentives in Nonprofit Organizations:

Evidence from Hospitals. Journal of Law and Economics 45(1): 227-249.

Brickley, J. A., Van Horn, R. L., and G. J. Wedig. (2010). Board Composition and Nonprofit

Conduct: Evidence from Hospitals. Journal of Economic Behavior & Organization 76(2):

196-208.

Buettner, R. (2011a). Reaping Millions in Nonprofit Care for the Disabled. New York Times

August 2: A1.

Callen, J. L., Klein, A., and D. Tinkelman. (2003). Board Composition, Committees, and

Organizational Efficiency: The Case of Nonprofits. Nonprofit and Voluntary Sector

Quarterly 32(4):493-520.

45

Cardinaels, E. (2009). Governance in Non-for-profit Hospitals: Effects of Board Remuneration

and Expertise on CEO Compensation. Health Policy 93(1): 64-75.

Carroll, T., Hughes, P., and W. Luksetich. (2005). Managers of Nonprofit Organizations Are

Rewarded for Performance. Nonprofit Management & Leadership 16(1): 19-41.

Chhaochharia, V., and Y. Grinstein. (2009). CEO Compensation and Board Structure. Journal of

Finance 64(1): 231-261.

Conyon, M. J., and L. He. (2011). Executive Compensation and Corporate Governance in China.

Journal of Corporate Finance 17(4): 1158-1175.

Core, J. E., Guay, W. R., and R. S. Verdi. (2006). Agency Problems of Excess Endowment

Holdings in Not-for-Profit Firms. Journal of Accounting and Economics 41(3): 307-333.

Core, J. E., Holthausen, R. W., and D. F. Larcker. (1999). Corporate Governance, Chief

Executive Officer Compensation, and Firm Performance. Journal of Financial

Economics 51(3): 371-406.

Cuomo, A. M. (2012, May 16). Governor Cuomo Announces Proposed Regulations to Ensure

that State-Funded Providers Do Not Pay Excessive Executive Compensation or

Administrative Costs [Press release]. The Office of the Governor of New York. Retrieved

from http://www.governor.ny.gov/press/05162012State-Funded-Providers

Dennison, M. (2011). Hospital Execs in Montana Still Getting the Big Bucks in 2009-10. The

Billings Gazette October 30.

Desai, M. A., and R. J. Yetman. (2006). Constraining Managers without Owners: Governance of

the Not-for-Profit Enterprise. NBER working paper No. 11140.

Doyle, J. (2011). Nonprofit Hospitals’ Huge Tax Breaks Under Increasing Scrutiny. St. Louis

Post-Dispatch October 23.

Eldenberg, L., Hermalin, B. E., Weisbach, M. S., and M. Wosinska. (2004). Governance,

Performance Objectives and Organizational Form: Evidence from Hospitals. Journal of

Corporate Finance 10(4): 527-548.

Evans, D. (2012, November 13). Tax-Exempt Firm Gets $600 Million Profit Flying First Class.

Bloomberg Markets Magazine.

Faleye, O., Reis, E., and A. Venkateswaran. (2013). The Determinants and Effects of CEO–

Employee Pay Ratios. Journal of Banking & Finance 37(8): 3258-3272.

Fama, E. F., and M. Jensen. (1983). Separation of Ownership and Control. Journal of Law and

Economics 26(2): 301-325.

46

Fries, A. (2013, June 9). Big Pay at Local Nonprofits. The Observer-Dispatch (Utica, New

York).

Froelich, K. A., and T. W. Knoepfle. (1996). Internal Revenue Service 990 Data: Fact or

Fiction? Nonprofit and Voluntary Sector Quarterly 25(1): 40-52.

Froelich, K. A., Knoepfle, T. W., and T. H. Pollak. (2000). Financial Measures in Nonprofit

Organization Research: Comparing IRS 990 Return and Audited Financial Statement

Data. Nonprofit and Voluntary Sector Quarterly 29(2): 232-254.

Frumkin, P., and E. K. Keating. (2010). The Price of Doing Good: Executive Compensation in

Nonprofit Organizations. Policy and Society 29(3): 269-282.

Glaeser, E. L. (2003). “Introduction.” In E. L. Glaeser (ed.), The Governance of Not-for-Profit

Firms. Chicago: University of Chicago Press.

Gose, B., and M. López-Rivera. (2012, September 16). Executive Pay Increased by Median of

3.8% in 2011, Chronicle Survey Finds. The Chronicle of Philanthropy. Retrieved from

http://www.philanthrophy.com/article/executive-pay-increased-by/134476/

Hallock, K. F. (2002). Managerial Pay and Governance in American Nonprofits. Industrial

Relations 41(3): 377-406.

Hancock, J. (2013, June 25). Grassley: Who Approved these Hospital Bonuses? The Washington

Post.

Hansmann, H. B. (1980). The Role of Nonprofit Enterprise. The Yale Law Journal 89(5): 835-

901.

Harford, J. (1999). Cash Reserves and Acquisitions. The Journal of Finance 54(6): 1969-1997.

Hopkins, B. R. (2009). Starting a Nonprofit Organization: A Legal Guide. 5th edition. New

Jersey: Wiley.

Internal Revenue Service. (2008). Return of Organization Exempt from Income Tax. Retrieved

from http://www.irs.gov/pub/irs-prior/f990--2008.pdf

Internal Revenue Service. (2007, June 14). IRS Releases Discussion Draft of Redesigned Form

990 for Tax-Exempt Organizations. Retrieved from http://www.irs.gov/uac/IRS-

Releases-Discussion-Draft-of-Redesigned-Form-990-for-Tax-Exempt-Organizations

Jensen, M. C., and W. H. Meckling. (1976). The Theory of the Firm: Managerial Behavior,

Agency Cost, and Ownership Structure. Journal of Financial Economics 3(4): 305-360.

Jensen, M. C., and K. J. Murphy. (1990). Performance Pay and Top-Management Incentives.

Journal of Political Economy 98(2): 225-264.

47

Jenter, D., and F. Kanaan. (2013). CEO Turnover and Relative Performance Evaluation.

Forthcoming in The Journal of Finance.

Kaplan, S. N., and B. A. Minton. (2006). How has CEO Turnover Changed? Increasingly

Performance Sensitive Boards and Increasingly Uneasy CEOs. NBER working paper No.

12465.

Keating, E. K., Parsons, L. M., and A. A. Roberts. (2008). Misreporting Fundraising: How Do

Nonprofit Organizations Accounts for Telemarketing Campaigns? The Accounting

Review 83(2): 417-446.

Krishnan, R., and M. H. Yetman. (2011). Institutional Drivers of Reporting Decisions in

Nonprofit Hospitals. Journal of Accounting Research 49(4): 1001-1039.

Krishnan, R., Yetman, M. H., and R. J. Yetman. (2006). Expense Reporting in Nonprofit

Organizations. The Accounting Review 81(2): 399-420.

Kruesi, K. (2011, November 3). Salaries and Expenses Rise at St. Luke’s. Times-News.

Malmendier, U., and G. Tate. (2009). Superstar CEOs. The Quarterly Journal of Economics

124(4): 1593-1638.

Miller, S. T. (2008, April 24). Remarks at the Representing & Managing Nonprofit

Organizations Conference, Georgetown University Law Center, Washington, D.C.

O’Regan, K., and S. M. Oster. (2005). Does the Structure and Composition of the Board Matter?

The Case of Nonprofit Organizations. The Journal of Law, Economics, & Organization

21(1): 205-227.

Opler, T., Pinkowitz, L., Stulz, R., and R. Williamson. (1999). The Determinants and

Implications of Corporate Cash Holdings. Journal of Financial Economics 52(1): 3-46.

Pakroo, P. H. (2011). Starting & Building a Nonprofit: A Practical Guide. 4th edition. Berkeley,

California: NOLO Press.

Pogrebin, R., and K. Taylor. (2010, April 26). Pulling the Reins (a Bit) on Hefty Salaries for

Cultural Executives. The New York Times: C4.

Roberts, A. A. (2005). The Implications of Joint Cost Standards for Charity Reporting.

Accounting Horizons 19(1): 11-23.

Roeger, K. L., Blackwood, A. S., and S. L. Pettijohn. (2013). The Nonprofit Almanac 2012.

Washington, D.C.: Urban Institute Press.

48

Roomkin, M. J., and B. A. Weisbrod. (1999). Managerial Compensation and Incentives in For-

Profit and Nonprofit Hospitals. Journal of Law, Economics, and Organization 15(3):

750-781.

Ryan, Jr., H. E., and R. A. Wiggins III. (2004). Who is in Whose Pocket? Director

Compensation, Board Independence, and Barriers to Effective Monitoring. Journal of

Financial Economics 73(3): 497-524.

Sataline, S. (2010). Illinois High Court Rules Nonprofit Hospital Can Be Taxed. The Wall Street

Journal March 19: B4.

Sedatole, K. L., Swaney, A., Yetman, M. H., and R. J. Yetman. (2012). Accounting-Based

Performance Metrics and Executive Compensation in Nonprofit Organizations. Working

paper.

Smith, E. B., and P. Kuntz. (2013, May 2). Disclosed: The Pay Gap Between CEOs and

Employees. Bloomberg Businessweek.

Steinberg, R. (Ed.). (2004). The Economics of Nonprofit Enterprises. Northampton, MA: Edward

Elgar Publishing.

Thornton, J. P., and W. H. Belski. (2010). Financial Reporting Quality and Price Competition

among Nonprofit Firms. Applied Economics 42: 2699-2713.

U.S. General Accounting Office. (2002). Tax-Exempt Organizations: Improvements Possible in

Public, IRS, and State Oversight of Charities. Retrieved from

http://www.gao.gov/new.items/d02526.pdf

Yetman, M. H., and R. J. Yetman. (2012). The Effects of Governance on the Accuracy of

Charitable Expenses Reported by Nonprofit Organizations. Contemporary Accounting

Research 29(3): 738-767.

48

Appendix A: Variable Definitions

See Panel 1 for definitions of the main variables and Panel 2 for definitions of governance index variables.

Numbers in brackets denote the corresponding Revised Form 990 section, line, and column numbers.

Panel 1: Main variables.

Access to Debt Set equal to 1 if the nonprofit utilized Tax-Exempt Bonds [X.20.B] during the ten-year period

ending in year t

Age The Age of the firm, where Age = 2012 – Year of Formation [Page 1, L]

Agency Dummy Set equal to 1 for cases of positive Excess Liquid Assets, and is 0 otherwise

Board Size The number of voting members of the governing body [VI.A.1a]

Capital Expenditures The ratio of Land, Buildings, and Equipment, Net [X.10c.B] to Total Assets [X.16.B]

CEO Tenure The number of years the CEO has held the title of CEO excluding the year of his or her

appointment

Commercial Dummy Set equal to 1 when the ratio of Program Services Revenue [VIII.2g.A] to Total Revenue

[VIII.12.A] is at least 90%, and 0 otherwise

CV(Total Revenue) The ratio of the standard deviation of Total Revenue [VIII.12.A] to mean Total Revenue

[VIII.12.A], where each is measured over the five-year period ending in year t

Donations Growth The change in the ratio of Net Contributions to Total Revenue [VIII.12.A] in year t–1 relative to year t–2 , where Net Contributions = Total Contributions [VIII.1h] – Government Revenue

[VIII.1e]

Excess Compensation The residual from a regression of Log(Relative Pay)it on the Log(No. of Employees)it, Log(Total

Assets)it, Log(Age)it–1, Log(CEO Tenure)it, Gender Dummy it, and Commercial Dummy it, as well

as industry fixed effects (based on the 26 NTEE major groupings) and year fixed effects

Excess Liquid Assets The residual from a fiscal year-specific regression of Liquid Assets/Total Expenses on Log(Total

Revenue), CV(Total Revenue), Access to Debt, and CV(Total Revenue) x Access to Debt, as well as industry fixed effects (based on the 26 NTEE major groupings) and state fixed effects

First-class Flights

Indicator

Set to 1 if the organization indicates that first-class or charter travel was provided to an

executive during year t [Schedule J, I.1a]

Fundraising Dummy Set equal to 1 when Total Fundraising Expenses [IX.25.D] are zero, and is 0 otherwise

Fundraising Ratio The ratio of (Net Contributions – Net Fundraising Expenses) to Net Contributions, where Net

Fundraising Expenses = Total Fundraising Expenses – Compensation allocated to Fundraising

Expenses ([IX.25.D] – [IX.5.D] – [IX.6.D] – [IX.7.D]) and Net Contributions = Total

Contributions – Government Revenue ([VIII.1h] – [VIII.1e])

Gender Dummy Set equal to 1 if the CEO is a female, and is 0 otherwise

Governance Index See “Governance Index” in Appendix A, Panel 2; defined as the ratio of the sum of the indicator

variables as a proportion of the total number of possible responses for each firm-year

observation, where each response is weighted by its annual cross-sectional standard deviation.

Government Revenue The ratio of Government Grants [VII.1e] to Total Revenue [VIII.12.A]

High Program Spending

Dummy

Set equal to 1 when the Program Spending Ratio exceeds the annual industry median value, and

is 0 otherwise

Household Income Median household income (as reported by the U.S. Census Bureau’s American Community Survey); computed for each state and fiscal year pair

Liquid Assets/Total

Expenses

The ratio of Liquid Assets to Net Total Expenses, where Liquid Assets = Cash + Savings and

Temporary Cash Investments + Pledges and Grants Receivable, Net + Accounts Receivable,

Net, and Net Total Expenses = Total Expenses – Compensation Expense

([X.1.B] + [X.2.B] + [X.3.B] + [X.4.B]) / ([IX.25.A] – [IX.5.A] – [IX.6.A] – [IX.7.A])

No. of Employees The Number of Employees [I.5]

NTEE-10 Industry

Classifications

A set of 10 groups defined by the National Taxonomy of Exempt Entities (NTEE): Arts,

Culture, and Humanities; Education; Environment and Animals; Health; Human Services;

International, Foreign Affairs; Public, Societal Benefit; Religion Related; Mutual/Membership

Benefit; and Unknown, Unclassified

NTEE-26 Industry

Classifications

A set of 26 groups defined by the National Taxonomy of Exempt Entities (NTEE): Animal-

Related; Arts, Culture, and Humanities; Civil Rights, Social Action, Advocacy; Community

Improvement, Capacity Building; Crime, Legal Related; Diseases, Disorders, Medical

Disciplines; Education; Employment, Job Related; Environmental Quality, Protection, and

Beautification; Food, Agriculture, and Nutrition; Health; Housing, Shelter; Human Services; International, Foreign Affairs, and National Security; Medical Research; Mental Health;

Mutual/Membership Benefit Organizations; Philanthropy, Voluntarism, and Grantmaking

49

Foundations; Public Safety; Public, Society Benefit; Recreation, Sports, Leisure, Athletics; Religion Related; Science and Technology Research Institutes; Social Science Research

Institutes; Youth Development; Unknown

Perquisites Indicator Set to 1 if the organization indicates that at least one type of perquisite was provided to an

executive during year t [Schedule J, I.1a] Note: The eight possibilities listed are first-class or

charter travel, travel for companions, tax indemnification and gross-up payments, discretionary

spending account, housing allowance or residence for personal use, payments for business use

of personal residence, health or social club dues or initiation fees, and personal services (e.g.,

maid, chauffeur, chef).

Program Expenses to

Assets

The ratio of Net Program Service Expenses to Total Assets, where Net Program Service

Expenses = Total Program Expenses – Compensation allocated to Program Service Expenses

([IX.25.B] – [IX.5.B] – [IX.6.B] – [IX.7.B] ) / [X.16.B]

Program Spending Ratio The ratio of Net Program Service Expenses to Net Total Expenses, where Net Program Service

Expenses = Total Program Expenses – Compensation allocated to Program Service Expenses

and Net Total Expenses = Total Expenses – Compensation Expense

([IX.25.B] – [IX.5.B] – [IX.6.B] – [IX.7.B] ) / ([IX.25.A] – [IX.5.A] – [IX.6.A] – [IX.7.A])

Relative Pay Ratio The ratio of CEO Compensation to Average Non-CEO Employee Pay (CEO Pay as identified in [VII.1a.D] divided by the sum of ([IX.5.A] + [IX.6.A] + [IX.7.A] –

CEO Compensation) scaled by the number of employees [I.5])

Revenue per Employee The ratio of Net Total Revenue to the Number of Employees [I.5], where Net Total Revenue =

Total Revenue [VIII.12.A] – Total Contributions [VIII.1h]

Total Assets Total Assets [X.16.B]

Total Revenue Total Revenue [VIII.12.A]

Total Liabilities/Total

Assets

The ratio of Total Liabilities [X.26.B] to Total Assets [X.16.B]

Panel 2: Components of the governance index.

Governing Body

Board Independence Is at least two-thirds of the organization’s board of directors is independent?

[VI.A.1b / VI.A.1a]

= 1 if Yes

= 1 if No

Inside Business

Relationship

Did any person who is a current or former officer, director, trustee, or key

employee have a direct business relationship with the organization [IV.28a],

have a family member who has a direct or indirect business relationship with

the organization [IV.28b], or serve as an officer, director, trustee, key

employee, partner, or member of an entity doing business with the

organization [IV.28c]?

= 1 if No

= 0 if Yes

Outside Management Did the organization delegate control over management duties customarily performed by or under the direct supervision of officers, directors or

trustees, or key employees to a management company or other person?

[VI.A.3]

= 1 if Yes

= 0 if No

Stockholders Does the organization have members or stockholders [VI.A.6] who may

elect one or more members of the governing body [VI.A.7a] and approve the

decisions of the governing body [VI.A.7b]?

= 1 if Yes

= 0 if No

Form 990 to Board Was a copy of the Form 990 provided to the organization’s governing body

before it was filed? [VI.A.10]

= 1 if Yes

= 0 if No

Governing Policies

Conflict of Interest Policy Does the organization have a written conflict of interest policy [VI.B.12a]

that is regularly and consistently monitored and enforced [VI.B.12c], and

that requires officers, directors or trustees, and key employees to disclose

annually interests that could give rise to conflicts [VI.B.12b]?

= 1 if Yes

= 0 if No

Whistleblower Policy Does the organization have a written whistleblower policy? [VI.B.13] = 1 if Yes

= 0 if No

Document Retention Policy

Does the organization have a written document retention and destruction policy? [VI.B.14]

= 1 if Yes = 0 if No

Grant to Officer Did the organization provide a grant or other assistance to an officer,

director, trustee, key employee, or substantial contributor, or to a person

related to such an individual? [IV.27]

= 1 if No

= 0 if Yes

50

Compensation Policies

CEO Compensation

Policy

Indicate the number of ways that the organization uses to establish the

compensation of the organization’s CEO/Executive Director [Schedule J,

I.3]. Note: The six possibilities listed are a compensation committee,

independent compensation consultant, Form 990 of other organizations,

written employment contract, compensation survey or study, and approval

by the board or compensation committee.

Continuous

variable

Non-CEO Compensation

Policy

Did the process for determining compensation of the non-CEO officers or

key employees of the organization include a review and approval by

independent persons, comparability data, and contemporaneous

substantiation of the deliberation and decision? [VI.B.15b]

= 1 if Yes

= 0 if No

Executive Reimbursement Policy

If the organization provided fringe benefits to any person listed in the Schedule of Compensation to Officers, Directors, Trustees, Key

Employees, and Highest Compensated Employees [Schedule J, I.1a], did

the organization follow a written policy regarding payment or

reimbursement or provision of all of the associated expenses [Schedule J,

I.1b]?

= 1 if Yes

= 0 if No

Executive Substantiation

Policy

If the organization provided fringe benefits to any person listed in the

Schedule of Compensation to Officers, Directors, Trustees, Key

Employees, and Highest Compensated Employees [Schedule J, I.1a], did

the organization require substantiation prior to reimbursing or allowing the

associated expenses [Schedule J, I.2]?

= 1 if Yes

= 0 if No

Accountability & Transparency

Tax Forms on Website Does the organization make its Form 1023 (or 1024 if applicable), 990,

and 990-T available for public inspection on its own website? [VI.C.18]

= 1 if Yes

= 0 if No

Audited Financial

Statements

Were the organization’s financial statements compiled, reviewed, or

audited by an independent accountant? [XI.2a-b] (If audited, the audit

must not be a Circular A-133 audit [i.e., “No” to XI.3b])

= 1 if Audited

= 0.50 if

Reviewed

or Compiled

= 0 if No

Audit Committee If the organization’s financial statements are compiled, reviewed, or

audited by an independent accountant [XI.2a-b], does the organization

have a committee that assumes responsibility for oversight of the audit,

review, or compilation of its financial statements and selection of an

independent accountant [XI.2c]?

= 1 if Yes

= 0 if No

a Industry classifications are based on the 26 NTEE major groupings.

51

Table 1: Descriptive statistics.

Panel A: Compensation variables.

Variable N Mean Std. Dev.

Lower

Quartile Median

Upper

Quartile

CEO Compensation 7,374 329,014 459,185 135,000 221,099 368,763

Relative Pay 7,374 11.330 10.381 4.655 8.053 14.437

Log(Relative Pay) 7,374 2.090 0.846 1.538 2.086 2.670

Panel B: Financial performance variables.

Program Spending Ratio 7,374 0.842 0.116 0.793 0.866 0.919

Industry-Adj. Program Spending Ratio 7,374 –0.019 0.113 –0.064 0.003 0.054

Program Expenses to Assets 7,374 0.414 0.509 0.126 0.268 0.491

Industry-Adj. Program Expenses to Assets 7,374 0.130 0.485 –0.096 0.000 0.176

Fundraising Ratio 6,582 0.902 0.176 0.883 0.961 1.000

Industry-Adj. Fundraising Ratio 6,582 –0.059 0.174 –0.068 0.000 0.015

Log(Revenue per Employee) 6,811 10.636 1.462 10.019 10.772 11.503

Industry-Adj. Log(Revenue per Employee) 6,811 –0.099 1.238 –0.539 0.012 0.450

Panel C: Determinants of relative pay and financial performance.

Governance Index 7,374 0.090 0.024 0.074 0.089 0.111

Board Size 7,374 21 13 12 17 26

No. of Employees 7,374 896 1,557 66 311 1,034

Total Assets 7,374 206,606,530 900,561,428 13,261,995 56,318,891 145,458,156

Total Revenue 7,374 93,469,039 258,346,966 6,884,526 22,869,960 76,881,128

Age 7,374 60 43 28 46 87

Liquid Assets/Total Expenses 7,374 0.850 1.217 0.281 0.489 0.906

Government Revenue 7,374 0.108 0.234 0.000 0.001 0.056

Donations Growth 7,201 0.007 0.227 –0.017 0.000 0.014

Capital Expenditures 7,374 0.366 0.263 0.112 0.379 0.562

Household Income 7,374 53,259 6,993 48,107 52,507 58,647

CEO Tenure 7,374 8.588 3.696 5.000 8.000 12.000

Gender Dummy 7,335 0.273 0.445 0.000 0.000 1.000

Commercial Dummy 7,374 0.378 0.485 0.000 0.000 1.000

Fundraising Dummy 7,374 0.352 0.478 0.000 0.000 1.000

Total Liabilities/Total Assets 7,374 0.404 0.371 0.141 0.347 0.585

Panel D: Number of organizations by NTEE industry classification.

Arts Education Health Human

Services Public

Benefit Other

No. of Observations 491 1,451 2,328 1,832 473 799

Descriptive statistics are computed for a maximum of 7,374 nonprofit-year observations covering the 2008 and 2009 tax years. Variable definitions are provided in Appendix A.

52

Table 2: Correlations between the CEO-to-employee relative pay ratio, financial performance, governance variables, and other determinants.

2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.

1. Log(Relative Pay) 0.01 –0.09 –0.07 –0.08 –0.02 0.27 0.69 0.45 0.36 –0.17 –0.17 –0.01 0.22 –0.01 0.02 –0.16 0.21 –0.06 0.12

2. Program Spending Ratio 0.29 0.11 0.05 0.04 0.03 0.08 0.06 0.01 –0.14 0.07 0.00 –0.03 0.02 0.05 0.02 0.07 0.03 0.06

3. Program Expenses to Assets 0.03 –0.10 0.03 –0.07 0.05 –0.25 –0.08 –0.20 0.26 –0.01 –0.11 0.03 –0.01 0.04 0.04 0.09 0.22

4. Fundraising Ratio –0.01 –0.01 0.03 –0.09 0.00 –0.06 0.05 0.00 0.05 –0.07 –0.06 –0.01 0.00 –0.14 0.29 –0.06

5. Log(Revenue per Employee) 0.03 –0.06 –0.10 0.30 –0.03 0.02 –0.52 –0.05 –0.12 0.03 –0.01 –0.07 0.25 0.13 0.07

6. Governance Index 0.01 0.01 0.04 –0.08 0.01 0.00 0.01 –0.10 0.09 0.05 0.04 0.01 –0.01 0.03

7. Board Size 0.18 0.22 0.33 0.02 –0.03 0.01 –0.04 –0.05 –0.03 –0.01 –0.23 –0.36 –0.14

8. Log(No. of Employees) 0.55 0.43 –0.32 –0.08 –0.05 0.30 0.00 –0.01 –0.19 0.42 0.04 0.24

9. Log(Total Assets) 0.29 –0.02 –0.27 0.01 –0.09 0.01 –0.03 –0.20 0.20 0.00 0.08

10. Log(Age) –0.18 –0.09 0.00 0.10 0.01 0.00 –0.09 0.05 –0.22 –0.10

11. Liquid Assets/Total Expenses 0.03 0.00 –0.20 0.04 0.01 0.05 –0.21 –0.05 –0.16

12. Government Revenue –0.04 0.03 0.00 0.02 0.09 –0.34 0.02 0.02

13. Donations Growth –0.07 0.02 0.01 0.01 –0.02 –0.01 –0.03

14. Capital Expenditures –0.04 0.00 –0.04 0.20 0.02 0.23

15. Log(Household Income) 0.06 0.03 0.02 –0.05 0.06

16. Log(CEO Tenure) 0.00 0.04 0.05 0.01

17. Gender Dummy –0.13 –0.08 –0.09

18. Commercial Dummy 0.37 0.31

19. Fundraising Dummy 0.26

20. Total Liabilities/Total Assets

Performance measures are industry adjusted. See Appendix A for variable definitions.

53

Table 3: Determinants of the CEO-to-employee relative pay ratio.

Performance measure: Program

Spending Ratio

Program Expenses

to Assets

Fundraising

Ratio

Log(Revenue per

Employee)

(I) (II) (III) (IV)

Industry-Adjusted Performance it –0.2601*** –0.1405*** –0.1251*** –0.0774***

(0.000) (0.000) (0.001) (0.000)

Governance Index it –0.5904** –0.5659** –0.3995 –0.2523

(0.040) (0.048) (0.162) (0.395)

Board Size it (÷ 10) 0.1060*** 0.0981*** 0.0930*** 0.0897***

(0.000) (0.000) (0.000) (0.000) Board Size Squared it (÷ 103) –0.0949*** –0.0870*** –0.0879*** –0.0802***

(0.000) (0.000) (0.000) (0.001)

Log(No. of Employees)it 0.3092*** 0.3243*** 0.3069*** 0.2793***

(0.000) (0.000) (0.000) (0.000)

Log(Total Assets)it 0.0362*** 0.0161** 0.0346*** 0.0735***

(0.000) (0.011) (0.000) (0.000)

Log(Age)it–1 (÷ 102) –0.0692 –0.6983 0.9836 0.2374

(0.946) (0.495) (0.329) (0.822)

Liquid Assets/Total Expenses it–1 (÷ 102) 1.1341* 0.5331 0.9511 0.1325

(0.070) (0.395) (0.132) (0.845)

Government Revenue it–1 –0.4225*** –0.3667*** –0.4172*** –0.5575***

(0.000) (0.000) (0.000) (0.000) Donations Growth it–1 (÷ 102) –0.0358 –0.1839 –4.1242

(0.994) (0.968) (0.415)

Capital Expenditures it–1 0.1068*** 0.0426 0.1215*** 0.1002***

(0.000) (0.170) (0.000) (0.001)

Log(Household Income) it–1 –0.1205** –0.1147** –0.1532*** –0.1316**

(0.022) (0.029) (0.003) (0.015)

Log(CEO Tenure)it 0.0783*** 0.0732*** 0.0765*** 0.0780***

(0.000) (0.000) (0.000) (0.000)

Gender Dummy it –0.0569*** –0.0598*** –0.0565*** –0.0571***

(0.000) (0.000) (0.000) (0.000)

Commercial Dummy it –0.1135*** –0.1000*** –0.1330*** –0.0650*** (0.000) (0.000) (0.000) (0.001)

Intercept 1.1931** 1.4834** 1.1531* 0.7799

(0.048) (0.014) (0.060) (0.209)

N 7,335 7,335 7,177 6,773

Adj. R2 0.555 0.559 0.558 0.557

Industry Fixed Effects? Yes Yes Yes Yes

Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Log(Relative Pay) on its traditional determinants, governance quality, and a

measure of financial performance: Program Spending Ratio in column I, Program Expenses to Assets in column II,

Fundraising Ratio in column III, and Log(Revenue per Employee) in column IV. Scaling of coefficients (where

applicable) is denoted in parentheses to the right of variable names. Donations Growth is excluded from column III to

avoid a potentially highly collinear relationship between the Fundraising Ratio and Donations Growth. See Appendix A

for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are

winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

54

Table 4: Determinants of nonprofit financial performance

Performance measure: Program

Spending Ratio

Program Expenses

to Assets

Fundraising

Ratio

Log(Revenue per

Employee)

(I) (II) (III) (IV)

Log(Relative Pay)it –0.0100*** –0.0789*** –0.0108*** –0.2064***

(0.000) (0.000) (0.003) (0.000)

Governance Index it 0.0789 0.1720 –0.0835 1.2081**

(0.158) (0.393) (0.344) (0.011)

Board Size it (÷ 102) 0.0259 –0.1864 –0.1977*** –0.4572

(0.443) (0.124) (0.000) (0.114) Board Size Squared it (÷ 104) 0.0149 0.2112 0.2960*** 0.2804

(0.748) (0.204) (0.000) (0.480)

Log(No. of Employees)it (÷ 10) 0.0472*** 1.2455*** –0.0094 –3.6444***

(0.001) (0.000) (0.684) (0.000)

Log(Total Assets)it (÷ 10) 0.0527*** –1.3745*** 0.0607*** 4.4859***

(0.000) (0.000) (0.002) (0.000)

Log(Age)it–1 (÷ 10) –0.0277 –0.2116*** –0.1032*** 0.4216**

(0.173) (0.004) (0.001) (0.014)

Liquid Assets/Total Expenses it–1 (÷ 10) –0.1179*** –0.5752*** –0.0227 –1.1231***

(0.000) (0.000) (0.244) (0.000)

Government Revenue it–1 0.0610*** 0.4331*** –0.0159 –1.9838***

(0.000) (0.000) (0.138) (0.000) Donations Growth it–1 (÷ 10) –0.0386 –0.0696 –4.6342***

(0.668) (0.830) (0.000)

Capital Expenditures it–1 –0.0301*** –0.5740*** –0.0062 –0.1505***

(0.000) (0.000) (0.517) (0.003)

Fundraising Dummy it–1 (÷ 10) 0.0202 0.4765*** 2.6807***

(0.550) (0.000) (0.000)

Total Liabilities/Total Assets it–1 (÷ 10) 0.1082** 3.8744*** 0.0039 3.3502***

(0.035) (0.000) (0.962) (0.000)

Commercial Dummy it 0.0248*** 0.0775*** –0.0465*** 0.5534***

(0.000) (0.000) (0.000) (0.000)

Intercept –0.0808** 2.0841*** 0.0104 –6.6862*** (0.029) (0.000) (0.887) (0.000)

N 7,374 7,374 7,220 6,811

Adj. R2 0.045 0.332 0.029 0.480

Industry Fixed Effects? Yes Yes Yes Yes

Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants,

governance quality, and (the log of) the CEO-to-employee relative pay ratio. Performance definitions considered include

the Program Spending Ratio in column I, Program Expenses to Assets in column II, Fundraising Ratio in column III, and

Log(Revenue per Employee) in column IV. Scaling of coefficients (where applicable) is denoted in parentheses to the right

of variable names. Donations Growth and the Fundraising Dummy are excluded from the Fundraising Ratio regression

(column III) to avoid a potentially spurious relationship between the dependent and independent variables. See Appendix

A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables

are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the

0.10, 0.05, 0.01 level for a two-tailed test.

55

Table 5: Determinants of the relative pay ratio at nonprofits – Instrumental variables estimations.

Performance measure: Program

Spending Ratio

Program Expenses

to Assets

Fundraising

Ratio

Log(Revenue

per Employee)

(I) (II) (III) (IV)

Industry-Adjusted Performance it –0.2641*** –0.1408*** –0.1993** –0.1044***

(endogenous) (0.002) (0.000) (0.017) (0.000)

Governance Index it –0.5801** –0.5677* –0.5157* –0.2604

(0.048) (0.051) (0.087) (0.380)

Board Size it (÷ 10) 0.0996*** 0.0907*** 0.0875*** 0.0867***

(0.000) (0.000) (0.000) (0.000) Board Size Squared it (÷ 103) –0.0881*** –0.0789*** –0.0760*** –0.0781***

(0.000) (0.001) (0.002) (0.001)

Log(No. of Employees)it 0.3111*** 0.3262*** 0.3153*** 0.2674***

(0.000) (0.000) (0.000) (0.000)

Log(Total Assets)it 0.0332*** 0.0128* 0.0267*** 0.0862***

(0.000) (0.054) (0.000) (0.000)

Log(Age)it–1 (÷ 102) –0.2936 –0.9332 0.5983 0.3166

(0.780) (0.374) (0.580) (0.764)

Liquid Assets/Total Expenses it–1 (÷ 102) 1.2463* 0.6533 1.3589** 0.0057

(0.053) (0.312) (0.049) (0.993)

Government Revenue it–1 –0.4254*** –0.3726*** –0.4413*** –0.6052***

(0.000) (0.000) (0.000) (0.000) Donations Growth it–1 (÷ 102) –0.4173 –0.4737 –6.1719

(0.929) (0.920) (0.228)

Capital Expenditures it–1 0.1009*** 0.0373 0.0956*** 0.0917***

(0.001) (0.249) (0.003) (0.003)

Log(Household Income) it–1 –0.1264** –0.1219** –0.1646*** –0.1316**

(0.018) (0.022) (0.003) (0.015)

Log(CEO Tenure)it 0.0772*** 0.0724*** 0.0718*** 0.0780***

(0.000) (0.000) (0.000) (0.000)

Gender Dummy it –0.0536*** –0.0572*** –0.0551*** –0.0587***

(0.001) (0.000) (0.001) (0.000)

Commercial Dummy it –0.1101*** –0.0977*** –0.1390*** –0.0469** (0.000) (0.000) (0.000) (0.020)

Intercept 1.3216** 1.6352*** 1.4440** 0.6028

(0.031) (0.007) (0.023) (0.334)

Sargan statistic (p-value) 2.312 (0.128) 1.212 (0.271) 0.001 (0.972) 0.190 (0.663)

Shea’s partial R2 (p-value) 0.523 (0.000) 0.733 (0.000) 0.245 (0.000) 0.419 (0.000)

N 7,003 7,003 6,185 6,743

Adj. R2 0.556 0.560 0.573 0.555

Industry Fixed Effects? Yes Yes Yes Yes

Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Log(Relative Pay) on its traditional determinants, governance quality, and a measure of industry-adjusted financial performance: Program Spending Ratio in column I, Program Expenses to Assets in

column II, Fundraising Ratio in column III, and Log(Revenue per Employee) in column IV. The regressions utilize a two-

stage least squares procedure where instruments for the first three measures of financial performance include the prior-period Total Liabilities/Total Assets ratio and the twice-lagged values of industry-adjusted financial performance. The only

exception is the Program Expenses to Assets specification, where the industry median Total Liabilities/Total Assets ratio is used in place of its firm-specific value to avoid over-identifying the model, given that both the endogenous and instrumental

variables share the same denominator. Due to an inability to identify the number of employees in t–2, I instrument for (Industry-Adjusted) Log(Revenue per Employee) using the contemporaneous median and twice-lagged values of Log(Total

Revenue). Scaling of coefficients (where applicable) is denoted in parentheses to the right of variable names. Donations Growth is excluded from column III to avoid a potentially highly collinear relationship between the Fundraising Ratio and

Donations Growth. Shea’s partial R2 and the Sargan statistic evaluate the quality of the instruments. See Appendix A for

variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are

winsorized at the 1st and 99

th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10,

0.05, 0.01 level for a two-tailed test.

56

Table 6: Determinants of nonprofit financial performance – Instrumental variables estimations.

Performance measure: Program

Spending Ratio

Program Expenses

to Assets

Fundraising

Ratio

Log(Revenue per

Employee)

(I) (II) (III) (IV)

Log(Relative Pay)it –0.0123*** –0.1170*** –0.0185*** –0.3094***

(endogenous) (0.001) (0.000) (0.002) (0.000)

Governance Index it 0.0745 0.0467 –0.1050 0.8910*

(0.191) (0.818) (0.255) (0.061)

Board Size it (÷ 102) 0.0419 –0.2074* –0.2372*** –0.4177

(0.224) (0.091) (0.000) (0.153) Board Size Squared it (÷ 104) –0.0001 0.2700 0.3437*** 0.2322

(0.999) (0.109) (0.000) (0.563)

Log(No. of Employees)it (÷ 10) 0.0521*** 1.3623*** 0.0332 –3.3458***

(0.002) (0.000) (0.246) (0.000)

Log(Total Assets)it (÷ 10) 0.0580*** –1.3755*** 0.0566*** 4.5464***

(0.000) (0.000) (0.005) (0.000)

Log(Age)it–1 (÷ 10) –0.0387* –0.2325*** –0.1034*** 0.2994*

(0.062) (0.002) (0.002) (0.085)

Liquid Assets/Total Expenses it–1 (÷ 10) –0.1141*** –0.5652*** –0.0147 –1.1240***

(0.000) (0.000) (0.482) (0.000)

Government Revenue it–1 0.0607*** 0.4000*** –0.0275** –2.0216***

(0.000) (0.000) (0.017) (0.000) Donations Growth it–1 (÷ 10) –0.0290 –0.0241 –4.7055***

(0.752) (0.941) (0.000)

Capital Expenditures it–1 –0.0294*** –0.5689*** –0.0098 –0.1696***

(0.000) (0.000) (0.334) (0.001)

Fundraising Dummy it–1 (÷ 10) 0.0197 0.4943*** 2.5084***

(0.568) (0.000) (0.000)

Total Liabilities/Total Assets it–1 (÷ 10) 0.0992* 3.9141*** 0.0080 3.4240***

(0.057) (0.000) (0.926) (0.000)

Commercial Dummy it 0.0248*** 0.0666*** –0.0491*** 0.5320***

(0.000) (0.000) (0.000) (0.000)

Intercept –0.0855** 2.1336*** 0.0204 –6.6131*** (0.021) (0.000) (0.780) (0.000)

Sargan statistic (p-value) 0.790 (0.374) 1.218 (0.270) 0.145 (0.703) 0.918 (0.338)

Shea’s partial R2 (p-value) 0.416 (0.000) 0.416 (0.000) 0.401 (0.000) 0.431 (0.000)

N 7,036 7,036 6,446 6,494

Adj. R2 0.044 0.330 0.029 0.485

Industry Fixed Effects? Yes Yes Yes Yes

Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants,

governance quality, and (the log of) the CEO-to-employee relative pay ratio. Performance definitions considered include

the Program Spending Ratio in column I, Program Expenses to Assets in column II, Fundraising Ratio in column III, and

Log(Revenue per Employee) in column IV. The regressions utilize a two-stage least squares procedure in which

instruments for the relative pay ratio include the contemporaneous value of the industry median of (the log of) the ratio of

CEO compensation to the number of volunteers serving the nonprofit, and (2) the twice-lagged value of (the log of) the

ratio of CEO compensation to total compensation paid by the organization. Shea’s partial R2 and the Sargan statistic

evaluate the quality of the instruments. Scaling of coefficients (where applicable) is denoted in parentheses to the right of variable names. Donations Growth and the Fundraising Dummy are excluded from the Fundraising Ratio regression

(column III) to avoid a potentially spurious relationship between the dependent and independent variables. See Appendix

A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables

are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the

0.10, 0.05, 0.01 level for a two-tailed test.

57

Table 7: Further analyses of the determinants of nonprofit financial performance.

Panel A: Changes specification Panel B: Lagged Log(Relative Pay)

Dependent variable:

Δ Industry-Adjusted

Program Spending Ratio it

Industry-Adjusted

Program Spending Ratio it

Δ Log(Relative Pay)it (÷ 102) –0.8375*** Log(Relative Pay)it–1 (÷ 102) –0.5855*

(0.007) (0.058)

Δ Governance Index it 0.1240 Governance Index it 0.0449

(0.118) (0.582)

Δ Board Size it (÷ 103) 0.1509 Board Size it (÷ 102) 0.0104 (0.741) (0.832)

Δ Board Size Squared it (÷ 106) 0.5227 Board Size Squared it (÷ 104) 0.0309

(0.850) (0.649)

Δ Log(No. of Employees)it (÷ 102) 0.2698 Log(No. of Employees)it (÷ 10) 0.0222

(0.596) (0.262)

Δ Log(Total Assets)it (÷ 102) 0.6422 Log(Total Assets)it (÷ 10) 0.0780***

(0.410) (0.000)

Δ Log(Age)it 0.0985** Log(Age)it–1 (÷ 10) –0.0294

(0.017) (0.301)

Δ Liquid Assets/Total Expenses it (÷ 102) –0.8648*** Liquid Assets/Total Expenses it–1 (÷ 10) –0.1060***

(0.000) (0.000) Δ Government Revenue it 0.0108 Government Revenue it–1 0.0674***

(0.428) (0.000)

Δ Donations Growth it (÷ 102) –0.3512 Donations Growth it–1 (÷ 10) –0.1170

(0.271) (0.287)

Δ Capital Expenditures it 0.0203 Capital Expenditures it–1 –0.0270***

(0.185) (0.001)

Δ Total Liabilities/Total Assets it (÷ 102) –0.5550 Fundraising Dummy it–1 (÷ 10) –0.0192

(0.451) (0.688)

Δ Fundraising Dummy it –0.0117* Total Liabilities/Total Assets it–1 (÷ 10) 0.1308*

(0.064) (0.066)

Δ Commercial Dummy it (÷ 102) 0.2490 Commercial Dummy it–1 0.0270*** (0.505) (0.000)

Intercept –0.0198 Intercept –0.5855*

(0.459) (0.058)

N 3,484 N 3,634

R2 0.021 Adj. R2 0.043

Industry Fixed Effects? Yes Industry Fixed Effects? Yes

Year Fixed Effects? Yes Year Fixed Effects? Yes

Panel A presents the results of pooled regressions where the change in the Industry-Adjusted Program Spending Ratio

is modeled as a function of changes in its traditional determinants, governance quality, and (the log of) the CEO-to-

employee relative pay ratio, where changes are computed as the difference between year t and year t–1 values. // Panel

B presents the results of pooled regressions where the Industry-Adjusted Program Spending Ratio is modeled as a

function of the lagged (rather than contemporaneous) value of (the log of) the CEO-to-employee relative pay ratio in

addition to the determinants introduced in comparable models. // Scaling of coefficients (where applicable) is denoted

in parentheses to the right of variable names. See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-

values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

58

Table 8: Executive perquisites, financial performance, and governance quality.

Logit Model

Perquisites

Indicatorit

Logit Model

First-class

Flight Indicatorit

Ordered Probit Model

Number of

Perquisitesit

(I) (II) (III)

High Program Spending Dummy it –11.8118*** –3.5954*** –4.2931***

(0.000) (0.000) (0.000)

Governance Index it –115.2116*** –44.8712*** –48.0889***

(0.000) (0.000) (0.000)

High Program Spending Dummy it x Governance Index it 17.3787*** 5.0814*** 6.2891***

(0.000) (0.000) (0.000)

Board Size it 0.0327*** –0.0479*** 0.0109**

(0.005) (0.001) (0.019)

Board Size Squared it –0.0002 0.0006*** –0.0001

(0.167) (0.001) (0.137)

Log(No. of Employees)it 0.3412*** 0.3029*** 0.1854***

(0.000) (0.000) (0.000)

Log(Total Assets)it 0.6521*** 0.6114*** 0.3103***

(0.000) (0.000) (0.000)

Log(Age)it–1 0.2013*** –0.2136** 0.0485*

(0.003) (0.027) (0.079)

Liquid Assets/Total Expenses it–1 –0.0880* –0.0898 –0.0605***

(0.059) (0.204) (0.004)

Government Revenue it–1 –1.2306*** –0.2172 –0.6603***

(0.000) (0.612) (0.000)

Donations Growth it–1 –0.1842 –0.1060 –0.0890

(0.568) (0.822) (0.520)

Capital Expenditures it–1 –0.9492*** –1.3723*** –0.4164***

(0.000) (0.000) (0.000)

Log(Household Income) it–1 0.4366 –0.9502* 0.0404

(0.244) (0.053) (0.777)

Log(CEO Tenure)it –0.1919** 0.1033 –0.0833**

(0.041) (0.395) (0.018)

Gender Dummy it –0.4536*** 0.0050 –0.2466***

(0.000) (0.975) (0.000)

Commercial Dummy it –0.1249 –0.1920 –0.0985**

(0.333) (0.231) (0.039)

Intercept –9.0616** 1.5044

(0.033) (0.787)

Cutpoint no. 1 2.9395*

(0.073)

Cutpoint no. 2 4.0937**

(0.013)

Cutpoint no. 3 4.7786***

(0.004)

Cutpoint no. 4 5.3997***

(0.001)

Cutpoint no. 5 5.9611***

(0.000)

Cutpoint no. 6 6.5602***

(0.000)

Cutpoint no. 7 7.5681***

(0.000)

N 5,767 5,732 5,767

Pseudo R2 0.601 0.276 0.338

Industry Fixed Effects? Yes Yes Yes

Year Fixed Effects? Yes Yes Yes

Column I models the probability of at least one perquisite being awarded to an executive; column II models the probability of first-class or

charter travel being awarded to an executive; and column III models the probability of multiple levels of executive perquisite types (0 to

8). Columns I and II represent multivariate logit models, while column III is a multivariate ordered probit model. Each specification is

modeled as a function of traditional determinants, governance quality, and an indicator variable for above industry median charitable

spending (High Program Spending Dummy). Of the 5,767 observations where disclosure of executive perquisites was required, 3,342

reported providing at least one type of perquisite and 372 provided first-class or charter travel. Of those nonprofits that provided at least

one perquisite, the mean (median) number of types provided was 1.97 (2.00). See Appendix A for variable definitions. Indust ry

classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99

th percentiles. p-

values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

59

Table 9: CEO Turnover, relative pay, and governance quality.

Panel A: A comparison of program spending ratios between turnover and (matched) non-turnover organizations.

Mean Program Spending Ratio Difference (Turnover – Non-Turnover)

Time relative to turnover (t = 0) Turnover Firms Non-Turnover Firms Difference p-value

Two years prior to turnover (t = –2) –0.035*** –0.026*** –0.001 0.249

One year prior to turnover (t = –1) –0.039*** –0.022*** –0.017*** 0.006

Year of turnover (t = 0) –0.036*** –0.020*** –0.016*** 0.001

One year after turnover (t = +1) –0.043*** –0.024*** –0.019** 0.050

Difference (t = +1 minus t = –1) –0.002 –0.004

p-value 0.761 0.233

Panel B: Multivariate logit regression predicting the probability of CEO turnover.

Log(Relative Pay) it–1 0.2528**

(0.041) High Program Spending Dummy it–1 0.4459

(0.207)

Log(Relative Pay) it–1 x High Program Spending Dummy it–1 –0.3044*

(0.067)

Log(CEO Tenure) it–1 –2.6845***

(0.000)

Gender Dummy it–1 –0.3634**

(0.044)

Log(Age) it–1 0.1438

(0.130)

Intercept –0.4720

(0.486)

N 4,659 Pseudo R2 0.378

Industry Fixed Effects? Yes

Year Fixed Effects? No

Panel A reports univariate analyses comparing the (Industry-Adjusted) Program Spending Ratios of turnover firms to a matched

sample of non-turnover firms, where a matching firm is required to have been in the same industry and of similar size (total assets,

with a differential of no more than 10%) in the year prior to the treatment firm’s turnover. // Panel B reports the results of a

multivariate logistic regression where the probability of CEO turnover is modeled as a function of the prior period relative pay

ratio, indicator variable for above industry median charitable spending (High Program Spending Dummy), and other relevant

CEO- and firm-specific determinants. Of the 4,659 observations, 270 entail CEO turnovers. See Appendix A for variable

definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st

and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-

tailed test.

60

Table 10: The relation between the relative pay ratio and governance quality for varying degrees of agency conflicts.

Dependent variable: Log(Relative Pay)

“Low Agency”

coefficients

Incremental effect

(× Agency Dummy)

(I) (II)

Board Independence it (÷ 10) 3.0635*** –3.2075*** (0.001) (0.010) Inside Business Relationship it (÷ 10) –0.0612 –0.4410 (0.908) (0.525) Outside Management it (÷ 10) 0.4524 –3.0870

(0.791) (0.139) Stockholders it (÷ 10) 0.1637 –0.1453 (0.845) (0.889) Form 990 to Board it (÷ 10) –0.2232 –0.6070 (0.721) (0.492) Conflict of Interest Policy it (÷ 10) 1.2880 –2.9629** (0.117) (0.012) Whistleblower Policy it (÷ 10) 0.3504 –1.8696*

(0.659) (0.080) Document Retention Policy it (÷ 10) –0.0962 1.5360 (0.899) (0.135) Grant to Officer it (÷ 10) 0.1068 0.1306 (0.912) (0.917) CEO Compensation Policy it (÷ 10) 1.7792* –2.1848* (0.061) (0.091) Non-CEO Compensation Policy it (÷ 10) –1.9424*** 1.8899** (0.002) (0.021)

Tax Forms on Website it (÷ 10) –2.1093*** 0.2151 (0.000) (0.790) Audited Financial Statements it (÷ 10) –2.6366 0.3972 (0.342) (0.919) Audit Committee it (÷ 10) 0.4960 0.8698 (0.685) (0.588) Board Size it (÷ 10) –0.0184 0.1035 (0.741) (0.146)

Board Size Squared it (÷ 1,000) 0.0637 –0.1420 (0.406) (0.146) Intercept 3.8342*** 0.3874 (0.008) (0.374)

N 1,407 Adj. R2 0.527 Industry Fixed Effects? Yes

Year Fixed Effects? Yes Control Variables? Yes

The results represent a pooled regression of Log(Relative Pay) on its traditional determinants, governance quality, the (Industry-Adjusted) Program Spending Ratio, and an indicator variable, Agency Dummy, to proxy for potentially severe agency conflicts. Agency Dummy is coded as 1 for the upper (lower) quartile of excess liquid asset levels, where excess liquid asset levels is defined to be the residual resulting from the model presented in equation 3. Column I reports the effects for “Low Agency” nonprofits (i.e., Agency Dummy = 0) while Column II provides the incremental effect for “High Agency” nonprofits (i.e., Agency Dummy = 1). To conserve space, only those coefficients corresponding to the governance variables are reported in this table. Scaling of coefficients (where

applicable) is denoted in parentheses to the right of variable names. Due to a relatively low response rate, the Executive Reimbursement Policy and Executive Substantiation Policy variables are excluded from the analysis. Control variables include the Industry-Adjusted Program Spending Ratioit, Board Sizeit, Board Size Squaredit, Log(No. of Employees)it, Log(Total Assets)it, Log(Age)it–1, Government Revenueit–1, Donations Growthit–1, Capital Expendituresit–1, Log(Household Income)it–1, Log(CEO Tenure)it, Gender Dummy it, and Commercial Dummyit . See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

61

Table 11: Economic significance of the impact of governance variables on the charitable spending of nonprofits with high agency conflicts versus nonprofits with low agency

conflicts.

Panel A: The impact on Log(Relative Pay) in response to switching answers to select governance questions.

Coefficient estimates (unscaled) Implied percentage change in Log(Relative Pay)

Low Agency group Increment High Agency group

One unit change in governance variables Low Agency group High Agency group

Governance variable (I) (II) (III) (IV) (V) (VI) Board Independence 0.3064 –0.3208 –0.0144 1.0000 35.85% –1.43% Conflict of Interest Policy 0.1288 –0.2963 –0.1675 1.0000 13.75% –15.42% Whistleblower Policy 0.0350 –0.1870 –0.1519 1.0000 3.57% –14.09% CEO Compensation Policy 0.1779 –0.2185 –0.0406 1.0000 19.47% –3.97% Non-CEO Compensation Policy –0.1942 0.1890 –0.0053 1.0000 –17.65% –0.52%

Panel B: The impact on Log(Relative Pay) in response to a one standard deviation increase in select governance variables.

Coefficient estimates (unscaled) Implied percentage change in Log(Relative Pay)

Low Agency group Increment High Agency group

One std. dev. change in governance variables Low Agency group High Agency group

Governance variable (I) (II) (III) (IV) (V) (VI) Board Independence 0.3064 –0.3208 –0.0144 0.2722 8.34% –0.39%

Conflict of Interest Policy 0.1288 –0.2963 –0.1675 0.2998 3.86% –5.02% Whistleblower Policy 0.0350 –0.1870 –0.1519 0.3555 1.25% –5.40% CEO Compensation Policy 0.1779 –0.2185 –0.0406 0.2594 4.62% –1.05% Non-CEO Compensation Policy –0.1942 0.1890 –0.0053 0.3489 –6.78% –0.18%

The results represent the economic significance of the governance variables whose statistical associations with the relative pay ratio are significantly related to the potential severity of agency conflicts (see Table 10). In Panel A, column V (VI) is calculated as ecoefficient – 1, where the “coefficient” value is given in column I (III). In Panel B, column V (VI) is computed as columns I × IV (III × IV).

62

Table 12: The relation between changes in charitable spending and changes in relative pay for varying degrees of

governance quality.

Δ Log(Relative Pay)it (÷ 102) –3.1355***

(0.009)

Governance Index it –0.0190

(0.696)

Δ Log(Relative Pay)it × Governance Index it 0.2491** (0.047)

Δ Board Size it (÷ 103) 0.1391

(0.760)

Δ Board Size Squared it (÷ 106) 0.5403

(0.845)

Δ Log(No. of Employees)it (÷ 102) 0.1268

(0.805)

Δ Log(Total Assets)it (÷ 102) 0.6768

(0.385)

Δ Log(Age)it 0.0997**

(0.016)

Δ Liquid Assets/Total Expenses it (÷ 102) –0.8522***

(0.000) Δ Government Revenue it 0.0107

(0.433)

Δ Donations Growth it (÷ 102) –0.3295

(0.301)

Δ Capital Expenditures it 0.0199

(0.192)

Δ Total Liabilities/Total Assets it (÷ 102) –0.5493

(0.456)

Δ Fundraising Dummy it –0.0121*

(0.056)

Δ Commercial Dummy it (÷ 102) 0.2595

(0.488)

Intercept –0.0174

(0.523)

N 3,484

R2 0.021

Industry Fixed Effects? Yes

Year Fixed Effects? Yes

The results represent pooled regressions where the change in the Industry-Adjusted Program Spending Ratio is modeled

as a function of changes in its traditional determinants and (the log of) the CEO-to-employee relative pay ratio, where changes are computed as the difference between year t and year t–1 values. The model includes the contemporaneous

value of the Governance Index to account for the impact of governance quality on the relation between changes in the

relative pay ratio and in charitable spending. Scaling of coefficients (where applicable) is denoted in parentheses to the

right of variable names. See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE

major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in

parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

63

Appendix B: Supplemental Evidence

Table B1: Determinants of the CEO-to-employee relative pay ratio for sub-samples of

commercial and traditional nonprofits.

Table B2: Determinants of nonprofit financial performance for sub-samples of

commercial and traditional nonprofits.

Table B3: Determinants of the CEO-to-employee relative pay ratio modeled as a function

of alternative measures of organizational size.

Table B4: Determinants of nonprofit financial performance modeled as a function of

alternative measures of organizational size.

Table B5: Determinants of nonprofit financial performance modeled as a function of

industry-adjusted relative pay.

Table B6: Determinants of nonprofit financial performance modeled as a function of

excess relative pay.

Table B7: Determinants of the CEO-to-employee relative pay ratio, modeled to include

state fixed effects.

Table B8: Determinants of nonprofit financial performance, modeled to include state fixed

effects.

Table B9: Estimation of excess liquid asset levels.

64

Table B1: Determinants of the CEO-to-employee relative pay ratio for sub-samples of commercial and traditional nonprofits.

Performance measure: Program Spending Ratio Program Expenses to Assets Fundraising Ratio Log(Revenue per Employee)

Commercial Traditional Commercial Traditional Commercial Traditional Commercial Traditional (I) (II) (III) (IV) (V) (VI) (VII) (VIII)

Industry-Adjusted Performance it –0.3638*** –0.1748** –0.1447*** –0.1195*** –0.0852* –0.1039* –0.2601*** –0.0550*** (0.001) (0.015) (0.000) (0.000) (0.079) (0.085) (0.000) (0.000)

Governance Index it 0.2203 –0.7238** 0.2141 –0.7103** –0.0560 –0.2761 0.5031 –0.3055 (0.645) (0.042) (0.653) (0.045) (0.907) (0.435) (0.285) (0.414)

Board Size it (÷ 10) 0.1017*** 0.0803*** 0.0990*** 0.0725*** 0.0667** 0.0814*** 0.0889*** 0.0645*** (0.002) (0.000) (0.002) (0.000) (0.045) (0.000) (0.006) (0.002) Board Size Squared it (÷ 103) –0.0839 –0.0630** –0.0822 –0.0551** –0.0751 –0.0685*** –0.0760 –0.0481* (0.100) (0.018) (0.106) (0.038) (0.136) (0.010) (0.129) (0.091) Log(No. of Employees)it 0.3081*** 0.3064*** 0.3264*** 0.3180*** 0.3307*** 0.3006*** 0.1861*** 0.2884*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Log(Total Assets)it 0.0965*** 0.0092 0.0700*** –0.0052 0.0832*** 0.0131* 0.2036*** 0.0372*** (0.000) (0.204) (0.000) (0.491) (0.000) (0.076) (0.000) (0.000) Log(Age)it–1 (÷ 102) –1.6470 2.0715 –2.6564 1.6201 0.2324 2.2528* –4.1833*** 3.0473**

(0.307) (0.117) (0.100) (0.220) (0.886) (0.079) (0.009) (0.028) Liquid Assets/Total Expenses it–1 (÷ 102) 0.8506 1.2742* –1.2921 0.8709 2.9478 0.7626 –3.9717** 0.7136 (0.644) (0.056) (0.491) (0.190) (0.130) (0.257) (0.038) (0.322) Government Revenue it–1 –0.1010 –0.4911*** –0.1077 –0.4363*** –0.2072 –0.4660*** –0.1202 –0.5845*** (0.590) (0.000) (0.565) (0.000) (0.295) (0.000) (0.512) (0.000) Donations Growth it–1 (÷ 102) –32.5026 0.6240 –30.2446 0.5565 –37.4517 –1.9579 (0.160) (0.893) (0.189) (0.905) (0.113) (0.702) Capital Expenditures it–1 –0.0320 0.1413*** –0.1111* 0.0912** 0.0137 0.1495*** –0.2018*** 0.1572***

(0.569) (0.000) (0.057) (0.014) (0.816) (0.000) (0.000) (0.000) Log(Household Income) it–1 –0.0585 –0.1642** –0.0541 –0.1617** –0.0473 –0.2096*** 0.0101 –0.1875*** (0.483) (0.016) (0.516) (0.017) (0.572) (0.002) (0.903) (0.008) Log(CEO Tenure)it 0.1274*** 0.0480*** 0.1184*** 0.0463*** 0.1360*** 0.0418*** 0.1174*** 0.0470*** (0.000) (0.004) (0.000) (0.005) (0.000) (0.008) (0.000) (0.007) Gender Dummy it –0.0403 –0.0607*** –0.0417 –0.0632*** –0.0222 –0.0663*** –0.0513* –0.0613*** (0.147) (0.001) (0.132) (0.001) (0.418) (0.000) (0.063) (0.002) Intercept –0.5381 2.0827*** –0.1078 2.3162*** 0.3537 2.2091*** –2.3712** 1.8863**

(0.578) (0.007) (0.911) (0.003) (0.742) (0.004) (0.014) (0.019)

N 2,774 4,561 2,774 4,561 2,570 4,607 2,727 4,046 Adj. R2 0.497 0.569 0.500 0.571 0.494 0.564 0.519 0.577 Industry Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes

The results represent pooled regressions of Log(Relative Pay) on its traditional determinants, governance quality, and financial performance for sub-samples of

commercial and traditional nonprofits, where commercial nonprofits are defined as those with at least 90% of revenues drawn from program services. Scaling of

coefficients (where applicable) is denoted in parentheses to the right of variable names. Donations Growth is excluded from column III to avoid a potentially highly

collinear relationship between the Fundraising Ratio and Donations Growth. See Appendix A for variable definitions. Industry classifications are based on the 26

NTEE major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at

the 0.10, 0.05, 0.01 level for a two-tailed test.

65

Table B2: Determinants of nonprofit financial performance for sub-samples of commercial and traditional nonprofits. Performance measure: Program Spending Ratio Program Expenses to Assets Fundraising Ratio Log(Revenue per Employee)

Commercial Traditional Commercial Traditional Commercial Traditional Commercial Traditional (I) (II) (III) (IV) (V) (VI) (VII) (VIII)

Log(Relative Pay)it –0.0116*** –0.0079*** –0.0899*** –0.0621*** –0.0163** –0.0048 –0.2217*** –0.2042*** (0.001) (0.009) (0.000) (0.000) (0.042) (0.179) (0.000) (0.000)

Governance Index it 0.2665*** –0.0218 0.3222 0.1595 –0.2979 –0.0046 1.5113*** 1.0911 (0.002) (0.766) (0.336) (0.517) (0.128) (0.957) (0.000) (0.123) Board Size it (÷ 102) –0.0726 0.0673 –0.2655 –0.1291 –0.2808** –0.1484*** –0.1527 –0.4144 (0.228) (0.109) (0.256) (0.360) (0.037) (0.002) (0.608) (0.314)

Board Size Squared it (÷ 104) 0.1907** –0.0416 0.5117 0.1101 0.5035** 0.2273*** 0.1942 0.0654 (0.040) (0.457) (0.155) (0.558) (0.014) (0.000) (0.671) (0.905) Log(No. of Employees)it (÷ 10) 0.0589** 0.0355** 1.5671*** 1.0236*** –0.0513 –0.0220 –3.9518*** –3.6237*** (0.019) (0.047) (0.000) (0.000) (0.433) (0.304) (0.000) (0.000) Log(Total Assets)it (÷ 10) 0.0219 0.0759*** –1.8212*** –1.1320*** 0.1033** 0.0400** 4.2429*** 4.8728*** (0.288) (0.000) (0.000) (0.000) (0.043) (0.027) (0.000) (0.000) Log(Age)it–1 (÷ 10) 0.0283 –0.0695** –0.3908*** –0.1664* –0.1480** –0.0860*** –0.7509*** 1.3121*** (0.342) (0.012) (0.001) (0.073) (0.026) (0.007) (0.000) (0.000)

Liquid Assets/Total Expenses it–1 (÷ 10) –0.0255 –0.1325*** –1.4769*** –0.4669*** 0.0240 –0.0260 –1.7568*** –0.9688*** (0.439) (0.000) (0.000) (0.000) (0.761) (0.113) (0.000) (0.000) Government Revenue it–1 0.0228 0.0656*** 0.0299 0.4540*** –0.1821** 0.0022 –0.0150 –2.0361*** (0.501) (0.000) (0.821) (0.000) (0.024) (0.812) (0.928) (0.000) Donations Growth it–1 (÷ 10) –1.1288*** 0.0112 –1.2733 0.0746 –1.9261 –4.4305*** (0.007) (0.907) (0.431) (0.817) (0.367) (0.000) Capital Expenditures it–1 –0.0446*** –0.0231*** –0.7602*** –0.4936*** –0.0203 –0.0013 –0.7241*** 0.1435** (0.000) (0.002) (0.000) (0.000) (0.417) (0.880) (0.000) (0.042) Fundraising Dummy it–1 (÷ 10) 0.0347 –0.0089 –0.0034 0.8536*** 0.7649*** 4.0156***

(0.452) (0.856) (0.985) (0.000) (0.001) (0.000) Total Liabilities/Total Assets it–1 (÷ 10) 0.2042*** 0.0913 3.7583*** 4.3079*** 0.3825** –0.1852** 2.0426*** 4.1918*** (0.007) (0.193) (0.000) (0.000) (0.034) (0.026) (0.000) (0.000) Intercept –0.0400 –0.1085** 2.9828*** 1.6653*** –0.2982 0.0330 –4.6353*** –7.9317*** (0.508) (0.022) (0.000) (0.000) (0.191) (0.590) (0.000) (0.000)

N 2,787 4,587 2,787 4,587 2,585 4,635 2,740 4,071 Adj. R2 0.022 0.058 0.322 0.376 0.019 0.029 0.598 0.452

Industry Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants, governance quality, and (the log of) the CEO-to-

employee relative pay ratio for sub-samples of commercial and traditional nonprofits, where commercial nonprofits are defined as those with at least 90% of revenues

drawn from program services. Scaling of coefficients (where applicable) is denoted in parentheses to the right of variable names. Donations Growth and the

Fundraising Dummy are excluded from the Fundraising Ratio regression (column III) to avoid a potentially spurious relationship between the dependent and

independent variables. See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

66

Table B3: Determinants of the CEO-to-employee relative pay ratio modeled as a function of alternative measures of organizational size.

Performance measure: Program Spending Ratio Program Expenses to Assets Fundraising Ratio Log(Revenue per Employee)

(I) (II) (III) (IV) (V) (VI) (VII) (VIII)

Industry-Adjusted Performance it –0.2750*** –0.1502** –0.0714*** –0.0676*** –0.1712*** –0.1418*** –0.1043*** –0.0947***

(0.000) (0.021) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000)

Governance Index it 1.1042*** 0.4971 1.0764*** 0.5027 1.1375*** 0.6633** 1.3125*** 0.8312**

(0.000) (0.116) (0.000) (0.111) (0.000) (0.036) (0.000) (0.011)

Board Size it (÷ 10) 0.1443*** 0.1513*** 0.1402*** 0.1479*** 0.1151*** 0.1316*** 0.1280*** 0.1449***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Board Size Squared it (÷ 103) –0.1047*** –0.1127*** –0.1011*** –0.1096*** –0.0786*** –0.0952*** –0.0913*** –0.1084***

(0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) (0.000)

No. of Employees it (÷ 104) 5.5852*** 5.5722*** 5.4440*** 5.3253***

(0.000) (0.000) (0.000) (0.000)

No. of Employees Squared it (÷ 108) –4.5914*** –4.5645*** –4.4110*** –4.4427***

(0.000) (0.000) (0.000) (0.000)

Total Assets it (÷1010) 0.7066*** 0.6014*** 0.6187*** 1.0820***

(0.000) (0.000) (0.000) (0.000)

Total Assets Squared it (÷ 1020) –0.1519*** –0.1299*** –0.1357*** –0.2335***

(0.001) (0.006) (0.002) (0.000)

No. of Employees it (÷ 104) 2.0465*** 2.0501*** 1.9697*** 1.9885***

(0.000) (0.000) (0.000) (0.000)

1 / No. of Employees it –1.9125*** –1.9510*** –1.7698*** –1.5963***

(0.000) (0.000) (0.000) (0.000)

Total Assets it (÷ 1010) –0.4331*** –0.4682*** –0.4389*** –0.3615***

(0.000) (0.000) (0.000) (0.000)

1 / Total Assets it (× 106) –0.1849*** –0.1589*** –0.2754*** –0.2368***

(0.000) (0.000) (0.000) (0.000)

Log(Age)it–1 (÷ 10) 0.8091*** 1.0413*** 0.7912*** 1.0219*** 0.9534*** 1.1184*** 0.7467*** 1.0048***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Liquid Assets/Total Expenses it–1 –0.0263*** –0.0232*** –0.0298*** –0.0272*** –0.0290*** –0.0366*** –0.0317*** –0.0340***

(0.000) (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Government Revenue it–1 –0.2910*** –0.3677*** –0.2596*** –0.3343*** –0.3063*** –0.3444*** –0.5736*** –0.5980***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Donations Growth it–1 –0.0509 –0.0747 –0.0545 –0.0782 –0.0667 –0.1100**

(0.299) (0.137) (0.266) (0.120) (0.211) (0.046)

Capital Expenditures it–1 0.2984*** 0.2951*** 0.2796*** 0.2755*** 0.3274*** 0.3161*** 0.2180*** 0.2136***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Log(Household Income) it–1 –0.0808 –0.0574 –0.0790 –0.0550 –0.1208** –0.0943 –0.1046* –0.0744

(0.150) (0.318) (0.159) (0.339) (0.030) (0.101) (0.067) (0.207)

Log(CEO Tenure)it 0.0715*** 0.0650*** 0.0677*** 0.0624*** 0.0728*** 0.0719*** 0.0739*** 0.0663***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Gender Dummy it –0.1000*** –0.1321*** –0.1005*** –0.1323*** –0.1116*** –0.1385*** –0.1061*** –0.1366***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Commercial Dummy it 0.0113 0.0582*** 0.0177 0.0669*** –0.0179 0.0416** 0.0480** 0.0960***

(0.559) (0.003) (0.364) (0.001) (0.361) (0.038) (0.015) (0.000)

Intercept 2.1071*** 2.0580*** 2.1166*** 2.0544*** 2.0305*** 1.9602*** 2.4228*** 2.2461***

(0.001) (0.002) (0.001) (0.002) (0.002) (0.003) (0.000) (0.001)

N 7,857 7,857 7,857 7,857 7,701 7,701 7,257 7,257

Adj. R2 0.466 0.438 0.466 0.439 0.473 0.438 0.477 0.442

Industry Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes

Year Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes

The results represent pooled regressions of Log(Relative Pay) on its traditional determinants, governance quality, and a measure of financial performance after replacing the organizationa l size

67

determinants used in Table 3 (Log(No. of Employees) and Log(Total Assets)) with alternative measures of size: employees and its squared term, along with total assets and its squared term (columns I, III,

V, and VII); and (2) the number of employees and its reciprocal, along with total assets and its reciprocal (columns II, IV, VI, and VII). Scaling of coefficients (where applicable) is denoted in parentheses

to the right of variable names. Donations Growth is excluded from columns V and VI to avoid a potentially highly collinear relationship between the Fundraising Ratio and Donations Growth. See

Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99

th percentiles. p-values are shown in

parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

68

Table B4: Determinants of nonprofit financial performance modeled as a function of alternative measures of organizational size.

Performance measure: Program Spending Ratio Program Expenses to Assets Fundraising Ratio Log(Revenue per Employee)

(I) (II) (III) (IV) (V) (VI) (VII) (VIII)

Log(Relative Pay)it –0.0089*** –0.0050** –0.0475*** –0.0446*** –0.0125*** –0.0096*** –0.2943*** –0.2494***

(0.000) (0.011) (0.000) (0.000) (0.000) (0.002) (0.000) (0.000)

Governance Index it 0.1537*** 0.1531*** –0.0047 0.2468 –0.0439 –0.0484 1.7783*** 1.7092***

(0.005) (0.005) (0.982) (0.232) (0.604) (0.572) (0.001) (0.001)

Board Size it (÷ 10) 0.0017 0.0030 –0.0333*** –0.0198 –0.0166*** –0.0158*** 0.0274 0.0323

(0.611) (0.363) (0.008) (0.109) (0.001) (0.002) (0.382) (0.295)

Board Size Squared it (÷ 103) 0.0032 0.0017 0.0373** 0.0207 0.0250*** 0.0245*** –0.0587 –0.0567

(0.482) (0.706) (0.031) (0.222) (0.000) (0.000) (0.173) (0.180)

No. of Employees it (÷ 104) 0.1913*** 0.4236*** 0.1070*** 0.0240

(0.000) (0.000) (0.009) (0.922)

No. of Employees Squared it (÷ 108) –0.1798*** –0.1998* –0.1267*** –0.5424**

(0.000) (0.078) (0.005) (0.049)

Total Assets it (÷ 1010) 0.1126*** –1.0989*** 0.1991*** 4.0573***

(0.000) (0.000) (0.000) (0.000)

Total Assets Squared it (÷ 1020) –0.0221*** 0.2376*** –0.0397*** –0.8909***

(0.009) (0.000) (0.001) (0.000)

No. of Employees it (÷ 104) 0.0549*** 0.1667*** 0.0268 –0.0084

(0.000) (0.000) (0.108) (0.934)

1 / No. of Employees it 0.0118 –0.5478*** 0.0387 2.9015***

(0.504) (0.000) (0.218) (0.000)

Total Assets it (÷ 1010) 0.0268 –0.4712*** 0.0485** 1.3973***

(0.115) (0.000) (0.034) (0.000)

1 / Total Assets it (× 106) –0.0062 0.3713*** –0.0103 –0.5651***

(0.225) (0.000) (0.279) (0.000)

Log(Age)it–1 (÷ 10) –0.0191 –0.0027 –0.1008 –0.0404 –0.1021*** –0.0913*** 0.1069 0.2328

(0.323) (0.887) (0.177) (0.578) (0.001) (0.002) (0.559) (0.193)

Liquid Assets/Total Expenses it–1 –0.0115*** –0.0119*** –0.0866*** –0.0794*** –0.0015 –0.0019 –0.0179 –0.0437***

(0.000) (0.000) (0.000) (0.000) (0.414) (0.299) (0.115) (0.000)

Government Revenue it–1 0.0563*** 0.0570*** 0.6125*** 0.5661*** –0.0218** –0.0200** –2.5749*** –2.4487***

(0.000) (0.000) (0.000) (0.000) (0.030) (0.049) (0.000) (0.000)

Donations Growth it–1 –0.0010 –0.0018 –0.0397 –0.0435 –0.3539*** –0.3122***

(0.905) (0.839) (0.239) (0.186) (0.000) (0.000)

Capital Expenditures it–1 –0.0272*** –0.0278*** –0.4304*** –0.4093*** –0.0054 –0.0076 –0.5721*** –0.6158***

(0.000) (0.000) (0.000) (0.000) (0.546) (0.396) (0.000) (0.000)

Fundraising Dummy it–1 0.0210 0.0288 0.3031** 0.3630*** 3.3655*** 3.2553***

(0.522) (0.381) (0.017) (0.003) (0.000) (0.000)

Total Liabilities/Total Assets it–1 0.1016** 0.1218** 3.7853*** 3.7495*** –0.0191 –0.0013 3.8382*** 4.0617***

(0.039) (0.013) (0.000) (0.000) (0.808) (0.987) (0.000) (0.000)

Commercial Dummy it 0.0238*** 0.0263*** 0.1264*** 0.1294*** –0.0446*** –0.0430*** 0.4031*** 0.4206***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

Intercept 0.0126 0.0030 0.1034 0.0103 0.0903 0.0960 –0.1790 –0.3298

(0.699) (0.927) (0.412) (0.934) (0.177) (0.152) (0.560) (0.275)

N 7,897 7,897 7,897 7,897 7,747 7,747 7,296 7,296

Adj. R2 0.045 0.041 0.233 0.267 0.028 0.026 0.342 0.365

Industry Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes

Year Fixed Effects? Yes Yes Yes Yes Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants, governance quality, and (the log of) the CEO-to-employee relative pay ratio

after replacing the organizational size determinants used in Table 3 (Log(No. of Employees) and Log(Total Assets)) with alternative measures of size: employees and its squared term, along with total assets and its squared term (columns I, III, V, and VII); and (2) the number of employees and its reciprocal, along with total assets and its reciprocal (columns II, IV, VI, and VII).

69

Scaling of coefficients (where applicable) is denoted in parentheses to the right of variable names. Donations Growth and the Fundraising Dummy are excluded from the Fundraising Ratio regression (columns V and VI) to avoid a potentially spurious relationship between the dependent and independent variables. See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test. Performance measures multiplied by a factor of 100.

70

Table B5: Determinants of nonprofit financial performance modeled as a function of industry-adjusted relative pay.

Performance measure: Program Spending Ratio

Program Expenses to Assets

Fundraising Ratio

Log(Revenue per Employee)

(I) (II) (III) (IV)

Industry-Adj. Log(Relative Pay)it –0.0099*** –0.0751*** –0.0109*** –0.2116*** (0.000) (0.000) (0.002) (0.000)

Governance Index it 0.0788 0.1732 –0.0841 1.2012** (0.159) (0.390) (0.340) (0.011) Board Size it (÷ 102) 0.0250 –0.1966 –0.1987*** –0.4715 (0.459) (0.105) (0.000) (0.102)

Board Size Squared it (÷ 104) 0.0157 0.2208 0.2971*** 0.2965 (0.733) (0.184) (0.000) (0.454) Log(No. of Employees)it (÷ 10) 0.0466*** 1.2315*** –0.0094 –3.6344*** (0.001) (0.000) (0.683) (0.000) Log(Total Assets)it (÷ 10) 0.0526*** –1.3759*** 0.0610*** 4.4874*** (0.000) (0.000) (0.002) (0.000) Log(Age)it–1 (÷ 10) –0.0278 –0.2124*** –0.1033*** 0.4196** (0.172) (0.004) (0.001) (0.014)

Liquid Assets/Total Expenses it–1 (÷ 10) –0.1180*** –0.5761*** –0.0227 –1.1221*** (0.000) (0.000) (0.245) (0.000) Government Revenue it–1 0.0611*** 0.4354*** –0.0158 –1.9845*** (0.000) (0.000) (0.140) (0.000) Donations Growth it–1 (÷ 10) –0.0418 –0.0942 –4.7038*** (0.642) (0.771) (0.000) Capital Expenditures it–1 –0.0303*** –0.5755*** –0.0063 –0.1527*** (0.000) (0.000) (0.511) (0.002)

Fundraising Dummy it–1 (÷ 10) 0.0198 0.4724*** 2.6708*** (0.559) (0.000) (0.000) Total Liabilities/Total Assets it–1 (÷ 10) 0.1071** 3.8652*** 0.0028 3.3280*** (0.037) (0.000) (0.973) (0.000) Commercial Dummy it–1 0.0248*** 0.0783*** –0.0465*** 0.5535*** (0.000) (0.000) (0.000) (0.000) Intercept –0.1004*** 1.9365*** –0.0115 –7.1053*** (0.007) (0.000) (0.876) (0.000)

N 7374 7374 7220 6811 Adj. R2 0.045 0.331 0.029 0.481 Industry Fixed Effects? Yes Yes Yes Yes Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants,

governance quality, and (the log of) the CEO-to-employee relative pay ratio, where the relative pay ratio has been adjusted

for its industry median. Scaling of coefficients (where applicable) is denoted in parentheses to the right of variable names.

Donations Growth and the Fundraising Dummy are excluded from the Fundraising Ratio regression (column III) to avoid a

potentially spurious relationship between the dependent and independent variables. See Appendix A for variable definitions.

Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and

99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-

tailed test.

71

Table B6: Determinants of nonprofit financial performance modeled as a function of excess relative pay.

Performance measure: Program Spending Ratio

Program Expenses to Assets

Fundraising Ratio

Log(Revenue per Employee)

(I) (II) (III) (IV)

Excess Compensation it –0.0102*** –0.0788*** –0.0114*** –0.2101*** (0.000) (0.000) (0.002) (0.000)

Governance Index it 0.0805 0.1699 –0.0861 1.2136** (0.151) (0.400) (0.331) (0.010) Board Size it (÷ 102) 0.0244 –0.1907 –0.1974*** –0.4515 (0.470) (0.117) (0.000) (0.119)

Board Size Squared it (÷ 104) 0.0167 0.2129 0.2956*** 0.2672 (0.719) (0.202) (0.000) (0.502) Log(No. of Employees)it (÷ 10) 0.0165 1.0098*** –0.0410** –4.2619*** (0.180) (0.000) (0.044) (0.000) Log(Total Assets)it (÷ 10) 0.0475*** –1.4156*** 0.0545*** 4.3808*** (0.000) (0.000) (0.005) (0.000) Log(Age)it–1 (÷ 10) –0.0303 –0.2381*** –0.1087*** 0.3409** (0.137) (0.001) (0.001) (0.048)

Liquid Assets/Total Expenses it–1 (÷ 10) –0.1183*** –0.5750*** –0.0222 –1.1366*** (0.000) (0.000) (0.258) (0.000) Government Revenue it–1 0.0617*** 0.4349*** –0.0170 –1.9839*** (0.000) (0.000) (0.115) (0.000) Donations Growth it–1 (÷ 10) –0.0285 –0.0550 –4.6565*** (0.751) (0.865) (0.000) Capital Expenditures it–1 –0.0304*** –0.5753*** –0.0061 –0.1552*** (0.000) (0.000) (0.522) (0.002)

Fundraising Dummy it–1 (÷ 10) 0.0266 0.4794*** 2.6875*** (0.433) (0.000) (0.000) Total Liabilities/Total Assets it–1 (÷ 10) 0.1136** 3.8755*** 0.0054 3.3316*** (0.027) (0.000) (0.948) (0.000) Commercial Dummy it–1 0.0251*** 0.0829*** –0.0459*** 0.5708*** (0.000) (0.000) (0.000) (0.000) Intercept –0.0782** 2.1093*** 0.0152 –6.6217*** (0.034) (0.000) (0.836) (0.000)

N 7,335 7,335 7,177 6,773 Adj. R2 0.046 0.332 0.029 0.481 Industry Fixed Effects? Yes Yes Yes Yes Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants,

governance quality, and (the log of) the CEO-to-employee relative pay ratio measured in excess form, where Excess

Compensation is defined as the residual from a regression of Log(Relative Pay)it on the Log(No. of Employees)it,

Log(Total Assets)it, Log(Age)it–1, Log(CEO Tenure)it, Gender Dummy it, and Commercial Dummy it, as well as industry

fixed effects (based on the 26 NTEE major groupings) and year fixed effects. Scaling of coefficients (where applicable)

is denoted in parentheses to the right of variable names. Donations Growth and the Fundraising Dummy are excluded

from the Fundraising Ratio regression (column III) to avoid a potentially spurious relationship between the dependent

and independent variables. See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in

parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

72

Table B7: Determinants of the CEO-to-employee relative pay ratio, modeled to include state fixed effects.

Performance measure: Program Spending Ratio

Program Expenses to Assets

Fundraising Ratio

Log(Revenue per Employee)

(I) (II) (III) (IV)

Industry-Adjusted Performance it –0.2597*** –0.1419*** –0.1254*** –0.0785*** (0.000) (0.000) (0.001) (0.000)

Governance Index it –0.5356* –0.5220* –0.3512 –0.2061 (0.064) (0.070) (0.223) (0.490) Board Size it (÷ 10) 0.1034*** 0.0959*** 0.0884*** 0.0888*** (0.000) (0.000) (0.000) (0.000)

Board Size Squared it (÷ 103) –0.0938*** –0.0862*** –0.0844*** –0.0808*** (0.000) (0.000) (0.000) (0.001) Log(No. of Employees)it 0.3115*** 0.3268*** 0.3096*** 0.2805*** (0.000) (0.000) (0.000) (0.000) Log(Total Assets)it 0.0337*** 0.0130** 0.0336*** 0.0713*** (0.000) (0.041) (0.000) (0.000) Log(Age)it–1 (÷ 102) –0.0591 –0.6144 0.7885 0.1985 (0.954) (0.552) (0.440) (0.853)

Liquid Assets/Total Expenses it–1 (÷ 102) 1.1537* 0.5404 1.1018* 0.2194 (0.065) (0.389) (0.082) (0.746) Government Revenue it–1 –0.4292*** –0.3739*** –0.4170*** –0.5688*** (0.000) (0.000) (0.000) (0.000) Donations Growth it–1 (÷ 102) –1.0004 –1.1133 –4.9586 (0.827) (0.807) (0.326) Capital Expenditures it–1 0.0971*** 0.0313 0.1128*** 0.0894*** (0.001) (0.316) (0.000) (0.004)

Log(Household Income) it–1 –0.2329 –0.2059 –0.3245 –0.1424 (0.440) (0.493) (0.276) (0.644) Log(CEO Tenure)it 0.0789*** 0.0737*** 0.0773*** 0.0792*** (0.000) (0.000) (0.000) (0.000) Gender Dummy it –0.0503*** –0.0536*** –0.0494*** –0.0510*** (0.001) (0.000) (0.001) (0.001) Commercial Dummy it –0.1193*** –0.1052*** –0.1403*** –0.0694*** (0.000) (0.000) (0.000) (0.000) Intercept 2.2291 2.2908 2.9728 0.9124

(0.500) (0.487) (0.362) (0.787)

N 7,335 7,335 7,177 6,773 Adj. R2 0.560 0.563 0.562 0.561 Industry Fixed Effects? Yes Yes Yes Yes State Fixed Effects? Yes Yes Yes Yes Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Log(Relative Pay) on its traditional determinants, governance quality, and financial performance after including state fixed effects to account for mandatory governance provisions in some U.S.

states. Scaling of coefficients (where applicable) is denoted in parentheses to the right of variable names. Donations

Growth is excluded from column III to avoid a potentially highly collinear relationship between the Fundraising Ratio

and Donations Growth. See Appendix A for variable definitions. Industry classifications are based on the 26 NTEE

major groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are shown in

parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-tailed test.

73

Table B8: Determinants of nonprofit financial performance, modeled to include state fixed effects.

Performance measure: Program Spending Ratio

Program Expenses to Assets

Fundraising Ratio

Log(Revenue per Employee)

(I) (II) (III) (IV)

Log(Relative Pay)it –0.0099*** –0.0791*** –0.0114*** –0.2079*** (0.000) (0.000) (0.002) (0.000)

Governance Index it 0.0958* 0.1867 –0.0488 1.0182** (0.090) (0.358) (0.582) (0.033) Board Size it (÷ 102) 0.0299 –0.1290 –0.1964*** –0.5434* (0.380) (0.290) (0.000) (0.062)

Board Size Squared it (÷ 104) 0.0148 0.1527 0.2771*** 0.4128 (0.750) (0.360) (0.000) (0.302) Log(No. of Employees)it (÷ 10) 0.0457*** 1.2618*** –0.0022 –3.6367*** (0.001) (0.000) (0.926) (0.000) Log(Total Assets)it (÷ 10) 0.0562*** –1.4017*** 0.0599*** 4.4938*** (0.000) (0.000) (0.002) (0.000) Log(Age)it–1 (÷ 10) –0.0378* –0.1563** –0.0834*** 0.4020** (0.067) (0.035) (0.009) (0.021)

Liquid Assets/Total Expenses it–1 (÷ 10) –0.1185*** –0.5740*** –0.0199 –1.0871*** (0.000) (0.000) (0.307) (0.000) Government Revenue it–1 0.0640*** 0.4299*** –0.0178* –1.9929*** (0.000) (0.000) (0.098) (0.000) Donations Growth it–1 (÷ 10) –0.0448 –0.0548 –4.4912*** (0.618) (0.865) (0.000) Capital Expenditures it–1 –0.0295*** –0.5857*** –0.0111 –0.1407*** (0.000) (0.000) (0.249) (0.005)

Fundraising Dummy it–1 (÷ 10) 0.0124 0.4916*** 2.6763*** (0.718) (0.000) (0.000) Total Liabilities/Total Assets it–1 (÷ 10) 0.1149** 3.9382*** 0.0526 3.2286*** (0.027) (0.000) (0.527) (0.000) Commercial Dummy it 0.0241*** 0.0794*** –0.0452*** 0.5447*** (0.000) (0.000) (0.000) (0.000) Intercept –0.1569*** 1.9622*** 0.0385 –5.5495*** (0.005) (0.000) (0.692) (0.000)

N 7,374 7,374 7,220 6,811 Adj. R2 0.052 0.338 0.043 0.483 Industry Fixed Effects? Yes Yes Yes Yes State Fixed Effects? Yes Yes Yes Yes Year Fixed Effects? Yes Yes Yes Yes

The results represent pooled regressions of Industry-Adjusted Financial Performance on its traditional determinants,

governance quality, and (the log of) the CEO-to-employee relative pay after including state fixed effects to account for

mandatory governance provisions in some U.S. states. Scaling of coefficients (where applicable) is denoted in parentheses to

the right of variable names. Donations Growth and the Fundraising Dummy are excluded from the Fundraising Ratio

regression (column III) to avoid a potentially spurious relationship between the dependent and independent variables. See

Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major groupings. All continuous

variables are winsorized at the 1st and 99th percentiles. p-values are shown in parentheses. *, **, *** indicates significance at

the 0.10, 0.05, 0.01 level for a two-tailed test.

74

Table B9: Estimation of excess liquid asset levels.

CV(Total Revenue) 1.1967***

(0.000)

Log(Total Revenue) –0.0505***

(0.000)

Access to Debt 0.0399

(0.150)

CV(Total Revenue) x Access to Debt –0.2945***

(0.002) Intercept 1.2350***

(0.001)

N 6,867

Adj. R2 0.233

Industry Fixed Effects? Yes

Year Fixed Effects? Yes

State Fixed Effects? Yes

The results represent a pooled regression of Liquid Assets/Total Expenses (i.e., “liquid asset

levels”) on determinants as in Core, Guay, and Verdi (2006). Excess liquid assets levels are

defined to be the residual from this model when estimated on a fiscal year-specific basis. See

Appendix A for variable definitions. Industry classifications are based on the 26 NTEE major

groupings. All continuous variables are winsorized at the 1st and 99th percentiles. p-values are

shown in parentheses. *, **, *** indicates significance at the 0.10, 0.05, 0.01 level for a two-

tailed test.